Вы находитесь на странице: 1из 36

IEEE TRANSACTIONS ON SMART GRID

Cognitive Radio Network for the Smart Grid:


Experimental System Architecture, Control
Algorithms, Security, and Microgrid Testbed
Robert Caiming Qiu, Senior Member, IEEE, Zhen Hu, Zhe Chen, Nan Guo, Senior Member, IEEE,
Raghuram Ranganathan, Member, IEEE, Shujie Hou, and Gang Zheng

Index TermsCognitive radio, grid control, ICA, microgrid, robust PCA, security, smart grid, testbed.

I. INTRODUCTION

promising? Ultimately, artificial intelligence and machine


learning will become the brain of the cognitive radioarguably any machine. How to put cognition into the cognitive
radio network is paramount at this moment. Experimental
knowledge is essential to dispel the fictional notion about
cognitive radio. What is cognitive radio? Obviously, cognitive
radio includes dynamic spectrum access (DSA) as a special
case. Adaptive systems may be a good alternative acronym
for cognitive systemselusive for some peoplein todays
technology.
The ever increasing demand on remote sensing capabilities
directly conflicts with the accelerating awareness of loss of
spectrum allocation. The FCC recommends spectrum policy
that makes 500 megahertz (MHz) of spectrum newly available
for broadband within 10 years, of which 300 MHz should be
made available for mobile use within five years. The need
for dynamic spectrum access using cognitive radio [2] is real
and immediate. The following paper series [3], [4] attempts to
bring together wireless communications with remote sensing,
especially radio frequency tomography.
The ultimate goal is to have integrated sensing, communication and computation systems, for various applications including the smart grid.

IE
W EE
eb P
r
Ve oo
rs f
ion

AbstractThis paper systematically investigates the novel idea


of applying the next generation wireless technology, cognitive radio
network, for the smart grid. In particular, system architecture, algorithms, and hardware testbed are studied. A microgrid testbed
supporting both power flow and information flow is also proposed.
Control strategies and security considerations are discussed. Furthermore, the concept of independent component analysis (ICA) in
combination with the robust principal component analysis (PCA)
technique is employed to recover data from the simultaneous smart
meter wireless transmissions in the presence of strong wideband
interference. The performance illustrates the gain of bringing the
state of the art mathematics to smart grid.

N THE FIRST issue of these TRANSACTIONS [1], the editorial points out that In a nutshell, smart grid amounts to providing an Internet Protocol (IP) address to every device that is
connected to the electricity grid. The presence of an IP address
enables a two-way communication mechanism between consumers and providers. Roughly speaking, the smart grid will: 1)
reduce blackouts; 2) promote renewable energy usage; 3) give
families more control over their energy diet. The motivation of
this paper is to address items 2 and 3, by studying the interdependencies of critical infrastructure in power and communications. In particular, a microgrid testbed is studied.
The advent of cognitive radio [2] is inevitable. The Moores
law ruthlessly drives computing faster and cheaper. Software
defined radio reflects this trend. However, the combination
of field programmable gate array (FPGA) and digital signal
processor (DSP) hardware is still too expensive. Is the general purpose computing on graphics processor unit (GPGPU)
Manuscript received October 14, 2010; revised April 11, 2011; accepted June
05, 2011. This work was supported by the National Science Foundation under
Grants ECCS-0901420, ECCS-0821658, and ECCS-0622125, and by the Office of Naval Research through Contracts N00014-07-1-0529 and N00014-11-10006. Paper no. TSG-00166-2010
The authors are with the Cognitive Radio Institute, Department of Electrical and Computer Engineering, Center for Manufacturing Research,
Tennessee Technological University, Cookeville, TN 38505, USA (e-mail:
rqiu@tntech.edu; zhu21@students.tntech.edu; zchen42@students.tntech.edu;
nguo@tntech.edu;
raghu80@gmail.com;
shou42@students.tntech.edu;
gzheng42@students.tntech.edu.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2011.2160101

A. Secure Communications in Smart Grid

Among the five identified key technology areas in smart grid,


the implementation of integrated communications is a foundational need, according to [5]. The smart grid in the near future will be required to accommodate increased demands for improved quality and energy efficiency. Solar and wind farms are
joining in for power generation in a distributed fashion. Appliances will become smart and talk to the control centers for optimum operations. Monitoring, managing, and controlling will
be required at all levels. Prediction of electricity prices, weather,
and social/human activities will be taken into account for optimum control. The addition of these new elements will result in
continuously increasing complexity. In order for different subnetworks or elements to be integrated into the smart grid seamlessly, a communication backbone has to be developed prior to
adding various functions. Hence, the earlier the communication
backbone is determined, the lesser the complications that would
be faced later in building the grid.
Although there have been many discussions on how to build
the communication network for the smart grid, it is desirable
to adopt existing communication standards. However, different

1949-3053/$26.00 2011 IEEE

IEEE TRANSACTIONS ON SMART GRID

simulations cannot answer, but also reveal more practical problems for further research. The capability of cognitive radio enables the smart grid, in many aspects, including security. With
minimal modifications to software, a cognitive radio network
testbed can also be used for the smart grid. Recently, there is a
trend of using TV band white space for the smart grid. Cognitive
radio is the most suitable wireless communications technology
using the white space.
The major concerns on hardware platforms for cognitive
radio networks are computing power and response latency. Our
proposed architecture addresses these two concerns. Cognitive
radio introduces intelligence beyond software defined radio
(SDR), like detection and learning algorithms, which means
cognitive radio requires much more computing power than
SDR. This enhanced computing power and reduced response
latency will enable enhanced (distributed) security.
A critical requirement for the smart grid is to enable accurate recovery of the smart meter wireless transmissions at the
central node or access point (AP). One of the impending challenges in this objective is the robustness of the data recovery in
the presence of strong wideband interference, due to easy access of the wireless data to any attacker, and inadequacy of existing physical layer security measures. In this paper, a novel
approach of applying a complex independent component analysis (ICA) technique [13] to smart meter data recovery is proposed, in combination with the recently developed robust principal component analysis (PCA) algorithm [14] for interference
cancellation and security enhancement.
For the feasible implementation of the power system to support the research on the smart grid, a microgrid testbed is proposed together with the potential system model. The microgrid
testbed will include various distributed energy resources, different power loads or appliances, and control modules. Layered
and hybrid control strategy will be explored in the microgrid to
coordinate power generation and power consumption. In order
to secure the microgrid, information security for the information flow and autonomous recovery for the power flow are the
key issues. Besides, control or optimization with uncertainty is
taken into account, because the microgrid is a dynamic complex
system and many of the variables in the corresponding control
or optimization issue cannot be deterministic or known for sure.
Thus, robust control and stochastic control are explored. Robust
control can guarantee the worst case performance, which will
make the microgrid stable and reliable.

IE
W EE
eb P
r
Ve oo
rs f
ion

from generic communication requirements, end-to-end security


and reliability are even more critical in smart grid communications [6]. According to the Department of Energy (DoE), one
of the critical challenges and opportunities lying ahead is increasing resilience, robustness, and security of the energy infrastructure. Security in the smart grid may be generalized into
two main areasphysical-layer and cybersecurity. In contrast
to cybersecurity that has been addressed intensively, there is
a large room to improve physical-layer security, especially for
wireless communications. For example, traditional spread spectrum techniques enhance security at the physical layer.
On the other hand, latency for time-sensitive data in the smart
grid has to be carefully considered. According to [7], a tolerance
of a maximum latency of 6 cycles, or 100 ms, can be used as a
benchmark for transmitting time-sensitive data. In other words,
to deliver time-sensitive data, dedicated low-latency communication links are required. The overall smart grid communication
network is a combination of different subnetworks with each
aiming at a certain set of requirements. Cognitive radio in the
white spaces (e.g., unused local TV broadcast spectra) can be
adopted to implement the dedicated low-latency communication links.
B. Related Work

The idea of using cognitive radio in the smart grid appears to


be proposed in the literature, for the first time, in [8][11]. In
particular, one of the three objectives of the submitted proposal
[8] in 2009 is apply the proposed network testbed for the smart
grid. The two-page white paper [12] is undated and was not
brought to our attention (through Qius student) until June 2010.
A cognitive radio network testbed is being built at TTU. Reference [11] is the first paper that captures the overall picture of
this projectthat may last at least for another 34 years. The
scope and the philosophy of this project are reported there.
A lot of future applications are network-centric. The design
of this testbed has taken into account the following applications.
Cognitive networking: What is the cognitive radio?
Smart grid: Cognitive radio enhances the network securitythe central challenge in smart grid.
Radiofrequency tomography imaging: The low-cost network has revolutionized remote sensing, e.g., through-wall
imaging.
C. Main Contributions

In this paper, the application of the fifth generation (5G) wireless technology, namely cognitive radio network, to the next
generation energy grid, the smart grid, has been addressed systemically for the first time.
Since the smart grid is a newly developed research topic,
references and literature on using communication testbeds for
smart grid are very less. In the area of wireless communications, cognitive radio is an emerging technique. The essence
of cognitive radio is the ability of communicating over unused
frequency spectrum adaptively and intelligently. A testbed for
cognitive radio networks will not only answer the questions that

II. COMMUNICATION TESTBEDS FOR SMART GRID

A. Hardware Platforms for Cognitive Radio Networks

There have been some wireless network testbeds, such as the


open access research testbed for next-generation wireless networks (ORBIT) [15] and the wireless testbed developed by University of California, Riverside [16]. Some common features
of those wireless network testbeds are summarized as follows.
First, the nodes in the networks are developed based on computer central processing units (CPUs). Second, the nodes use

IE
W EE
eb P
r
Ve oo
rs f
ion

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 1. USRP2 with WBX RF daughterboard.

Fig. 2. SFF SDR DP with low-band tunable RF module.

802.11 Wi-Fi network interface cards for wireless communications. These network testbeds may work well for evaluating algorithms, protocols, and network performances for Wi-Fi networks. But they are not suitable for cognitive radio networks,
due to their inherent lack of wide-band frequency agility.
Recently, Virginia Tech developed a testbed for cognitive
radio networks with 48 nodes [17], which is an significant
achievement in this area. Each node consists of three parts:
an Intel Xeon processor-based high-performance server, a
Universal Software Radio Peripheral 2 (USRP2), and a custom
developed radiofrequency (RF) daughterboard that covers a
continuous frequency range from 100 MHz to 4 GHz with
variable instantaneous bandwidths from 10 kHz to 20 MHz.
The node is easily capable of frequency agility. However, as
the authors mentioned, the drawbacks of the node are twofold.
First, it is not a low-power processing platform. Second, it is
not capable of mobility.
Regardless of the kind of cognitive radio network testbed, it
is composed of multiple nodes. There exist some commercial
off-the-shelf hardware platforms designed for SDR that may be
used for building the nodes for cognitive radio networks.
1) Universal Software Radio Peripheral 2: USRP and
USRP2 provided by Ettus Research are widely used hardware
platforms in the area of SDR and cognitive radio. USRP2 is
the second generation of USRP, and it became available in
2009 [18]. USRP2 consists of a motherboard, and one or more
selectable RF daughterboards, as shown in Fig. 1.
The major computation power on the motherboard comes
from a Xilinx Spartan-3 XC3S2000 FPGA. The motherboard is
also equipped with a 100 MSPS 14-bit dual channel analog-todigital converter (ADC), a 400 MSPS 16-bit dual channel digital-to-analog converter (DAC), and a gigabit ethernet port that
can be connected to a host computer. There are some RF daughterboards available for USRP2. Among them, a newly developed RF daughterboard called WBX covers a wide frequency
band of 50 MHz to 2.2 GHz, with a nominal noise figure of 57
dB.
Signals are received and down-converted by USRP2, and
its RF daughterboard. Subsequently, they are sent to a host
computer for further processing through the gigabit ethernet.

Most of the processing work is done by the host computer. Data


to be transmitted are sent from the host computer to USRP2
through the same gigabit ethernet, before they are up-converted
and transmitted by USRP2 and its RF daughterboard.
A major advantage of USRP2 is that it works with GNU
radio [19], a open source software with plenty of resources for
SDR and a lot of users, which simplifies and eases the usage
of USRP2. On the other hand, USRP2 is not perfect. First,
the gigabit Ethernet connecting USRP2 and its host computer
introduces random time delays. The operating system on the
host computer may also introduce random time delays. According to our measurement, the response delay of USRP2 is in
the range of several milliseconds to tens of milliseconds [20].
Such random response delay may be acceptable for half-duplex communications. However, in cognitive radio networks,
full-duplex communications are desired and random response
delays may deteriorate the performance of cognitive radio
networks. Second, USRP2 is usually used together with GNU
radio that runs on a host computer. When the instantaneous
bandwidth of USRP2 increases, the CPU on the host computer
gets much busier. Therefore, a multicore CPU is desired,
similar to what Virginia Tech has done to their network testbed.
When the instantaneous bandwidth of USPR2 becomes wider,
and the processing tasks on GNU radio becomes much more
complex, a common CPU may not be competent enough for
real-time processing.
2) Small Form Factor Software Defined Radio Development
Platform: The small form factor (SFF) SDR development platform (DP) provided by Lyrtech in collaboration with Texas Instruments (TI) and Xilinx is a self-contained platform consisting
of three separate boards: digital processing module, data conversion module, and RF module, as shown in Fig. 2 [21][23].
The digital processing module is designed based on
TMS320DM6446 system-on-chip (SoC) from TI and Virtex-4
SX35 FPGA from Xilinx. The TMS320DM6446 SoC has
a C64x+ DSP core running at 594 MHz together with an
advanced RISC machine (ARM9) core running at 297 MHz.
The digital processing module also comes with a 10/100 Mbps
Ethernet port. The data conversion module is equipped with a
125 MSPS 14-bit dual channel ADC and a 500 MSPS 16-bit

IEEE TRANSACTIONS ON SMART GRID

IE
W EE
eb P
r
Ve oo
rs f
ion

Fig. 3. WARP FPGA board with two radio boards.

dual channel DAC. It also has a Xilinx Virtex-4 LX25 FPGA.


The low-band tunable RF module can be configured to have
either 5 MHz or 20 MHz bandwidth with working frequencies
of 2001050 MHz for the transmitter, and 2001000 MHz for
the receiver. The nominal noise figure of this RF module is 5
dB. Other frequency bands may be covered by several other
RF modules.
There are two favorable features of SFF SDR DP for cognitive radio networks. One is that SFF SDR DP is in small form
factor and can be moved easily. The other is that it is capable
of supporting full-duplex communications. However, there are
also two technical drawbacks of using it to build nodes for cognitive radio networks. One drawback is that its computing capacity is fixed, and it is not easy to upgrade to meet the demands
of cognitive radio networks. The other drawback is the response
time delay. According to our measurement, the response delay
of SFF SDR DP is about tens of milliseconds, and the delay is
constant [20]. Such a nontrivial delay is undesirable for cognitive radio networks, since it may deteriorate the performance.
SFF SDR DP can be viewed as an example of independent
hardware platforms, whereas USRP2 is an example of computer-aided hardware platforms. A comparison between the two
hardware platforms has been reported in [11].
3) Wireless Open-Access Research Platform: The wireless
open-access research platform (WARP) developed by Rice University consists of an FPGA board, and one to four radio boards
[24], as shown in Fig. 3. The second generation of the FPGA
board has a Xilinx Virtex-4 FX100 FPGA and a gigabit ethernet port [25], [26]. The FPGA can be used to implement the
physical layer of wireless communications. There are PowerPC
processors embedded in the FX100 FPGA that can be used to
implement media access control (MAC) and network layer. The
radio board incorporates a dual-channel 65 MSPS 14-bit ADC,
and a dual-channel 125 MSPS 16-bit DAC, covering two frequency ranges of 24002500 MHz and 49005875 MHz, with
a bandwidth of up to 40 MHz.
WARP platform is also a small form factor independent hardware platform, which is attractive for building the nodes of cognitive radio networks. The second advantage of using WARP is
that both the physical layer and MAC layer can be implemented

Fig. 4. Sora radio control board.

on one FPGA, which may simplify the board design, compared


to an FPGA + DSP/ARM architecture. Hence, time delays
introduced by the interface between FPGA and DSP/ARM can
be reduced. However, according to [26], the Virtex-4 FPGA on
WARP is not powerful enough to accommodate both transmitter
and receiver functions at the same time. Thus, full-duplex communications desired by cognitive radio networks cannot be implemented using just one WAPR platform.
4) Microsoft Research Software Radio: Microsoft research
has developed a Software radio (Sora) platform [27]. Sora is
composed of a radio control board (RCB), and a selectable RF
board, and it works with a multicore host computer. The RCB
is shown in Fig. 4.
The RCB contains a Xilinx Virtex-5 FPGA, and it interfaces
with a host computer through a peripheral component interconnect express (PCIe) interface at a rate of up to 16.7 Gbps. Actually, RCB is an interface board for transferring digital signals
between the RF board and computer memory. The RF board
can be a WARP radio board. Processing work including physical layer and MAC layer is done on the host computer.
Sora is a computer-aided platform. The main advantage of
using Sora is that it provides a high-throughput interface between RF boards and a host computer. However, since processing work burdens the host computer, the host computer has
to be very powerful to support all the functions running in real
time. On the other hand, multicore programming and debugging
with speedup tricks is not easy. Moreover, implementing fullduplex communications on one host computer is challenging.
Obviously, a host computer (or server) installed with Sora lacks
mobility.
B. Proposed Testbed for Cognitive Radio Networks and Smart
Grid

All of the above four hardware platforms are designed for


SDR. Two of them connect to a host computer where major processing work is done. The other two are stand-alone hardware
platforms. From the aspect of mobility, stand-alone platforms
are preferable for building the nodes of cognitive radio networks, whereas from the aspect of software development, computer-aided hardware platforms are more practical, since soft-

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 5. Architecture of the proposed motherboard for hardware platforms.

IE
W EE
eb P
r
Ve oo
rs f
ion

ware development and debugging on a host computer is generally easier. In [28], a compromise between the above two kinds
of hardware platforms is suggested. The authors recommend
performing time-critical tasks in the FPGA, and split MAC design with host and FPGA implementations.
However, to the best of our understanding, compared to the
hardware platforms for SDR, the major concerns on hardware
platforms for cognitive radio networks are computing power
and response time delay. Cognitive radio introduces intelligence beyond SDR, like detection and learning algorithms,
which means cognitive radio requires much more computing
power than SDR. A hardware platform with ample and upgradable computing power is desired for building cognitive radio
testbeds. On the other hand, the desired hardware platform
should have minimum response time delay. If the response time
delay is large, the throughput of cognitive radio networks will
seriously degrade. Moreover, full-duplex communications for
the desired hardware platforms is preferable.
Unfortunately, none of the existing off-the-shelf hardware
platforms can meet the above requirements at the same time.
They are originally designed for SDR, instead of cognitive radio
networks. It is imperative to design a new hardware platform for
building the nodes of cognitive radio networks.
1) Proposed Motherboard for Hardware Platforms: We propose an architecture for the motherboard of the proposed hardware platform. Regarding the RF front-end, existing RF boards
from WARP or USRP2 can be reused to interface with the proposed motherboard. Therefore, one proposed motherboard, and
one or more RF boards constitute one proposed hardware platform.
Fig. 5 shows the proposed architecture of the first generation motherboard and its major components. Two powerful
FPGAs, i.e., a Virtex-6 FPGA and a Virtex-5 FX FPGA, are
employed as core components on the motherboard. All the functions for physical layer and MAC layer are implemented on the
two FPGAs, and no external host computer is required. The proposed hardware platform is stand-alone, thus it has good mobility. The Virtex-5 FX FPGA has PowerPC cores that are dedicated for implementing the MAC layer. Physical layer functions
including spectrum sensing are implemented on the two FPGAs.
The Virtex-5 FPGA is used for the transmitting data path, and it
is connected to one or two RF boards as well as a gigabit ethernet
port. The Virtex-6 FPGA is dedicated for the receiving data
path, with connections to one or two RF boards and an extension
port. The extension port can be used to connect with external
boards to gain access to additional computing resources. The
two FPGAs are connected together by a high-throughput low-latency onboard bus. Both of the FPGAs have access to their own
external memories. The use of two FPGAs is a trade-off between
performance and cost.
The proposed motherboard can provide enough and upgradable computing resource for cognitive radio networks. In addition, the time delays between the two FPGAs are trivial. Moreover, full-duplex communications are supported by one proposed motherboard with two or more RF boards.
2) Proposed Functional Architecture for Building Nodes for
Network Testbeds: Based on the proposed motherboards and
off-the-shelf RF boards, nodes for network testbeds can be im-

Fig. 6. Proposed functional architecture for the nodes.

plemented using the following proposed functional architecture,


as shown in Fig. 6.
The hardware abstraction layer (HAL) is a packaged interface
for upper-level functions that screens hardware-specific details.
It provides data interfaces to both receiving data path and transmitting data path, as well as an access interface to other hardware-specific resources on the hardware platform. The spectrum
and channel manager manages all the spectrum and channel related resources, including links, frequencies, and modulation
methods. There are several functional modules interfaced with
the spectrum and channel manager. The spectrum detection and
prediction module provides the information regarding the availability of some frequency bands. The decision making module
utilizes decision algorithms to make decisions such as which
channel will be used, and when it will be used. More learning al-

IEEE TRANSACTIONS ON SMART GRID

the capability of protecting themselves from being invaded or


tampered. In [32], it is pointed out that FPGAs unfortunately
open another door for malicious users to implement the hardware analogue of a computer virus. Moreover, the protection
of the information processed within the FPGAs is discussed in
[33], [34]. For the smart grid, how do the nodes in the microgrids
prevent their information from being invaded or tampered? This
is an open question for further research.
III. HOST COMPUTER WITH GRAPHICS PROCESSING UNIT AS
COMPUTING ENGINE

IE
W EE
eb P
r
Ve oo
rs f
ion

The testbed capability can be greatly expanded if a computing


engine is attached. The computing engine can be a host computer which undertakes many tasks including high-computation
oriented tasks. Recently, a computing enhancement technology
called GPGPU emerged in the PC industry. GPGPU refers to a
relatively new method by which the various cores of a graphics
processor unit (GPU) can be utilized for general purpose parallel computing [35]. This idea of utilizing GPUs for nongraphical applications first became popular in 2003, but was limited
by the amount of knowledge required to successfully write such
programs. November 2008 saw the introduction of Nvidias G80
architecture which brought greater versatility through support
of the C computer language, and a more generalized and programmer friendly hardware structure [36].
CUDA is both a hardware and software architecture by
Nvidia, which is actually what allows GPUs to run programs
that have been written using C, C++, Fortran, etc. It works by
executing a kernel across several parallel threads [36]. GPUmat
allows standard MATLAB code to run on GPUs. Furthermore,
CULA is a linear algebra library which has been designed
to utilize the NVidia CUDA architecture for computational
acceleration. This library is designed in a manner such that
those with little or no GPGPU programming experience can
take advantage of the parallel computing power offered by
GPGPU. Given the flexibility offered by this product, it would
be advantageous to consider it for use with the test-bed.
The CULA library is compatible with Python, C/C++, Fortran, and Matlab. When using C/C++, the library is designed
in such a way that the user may simply replace existing functions in the program with those from the library. CULA is designed such that it automatically handles the memory allocation
required with GPGPU programming. This is the most attractive feature of this software because it allows users who are not
experienced with GPGPU programming to take advantage of
the increased speed it offers. The code is also flexible enough
so that more experienced programmers can manually adjust the
memory allocation for the GPU. Such features are very valuable
to our lab because they allow a diverse set of people to work with
the GPUs, while still retaining the option for more advanced
programming. CULA can also use the Matlab MEX compiler
to compile code from within the Matlab environment, which is
capable of utilizing the GPGPU parallel structure. From the descriptions given, this seems to be very similar to the open source
library GPUmat. Our lab has investigated GPUmat in the past
as a means for GPGPU programming, but it was felt that it still
lacked many of the functions we would require for our work.

Fig. 7. Proposed network testbed.

gorithms can be implemented as an independent module to learn


and reason from the inputs. The geolocation module outputs the
latitude and longitude of the node. The spectrum and channel
manager can use such geolocation information to load prior information about current location from the knowledge/policy/
data base. The routing manager employs routing algorithms to
select the best route for sending and relaying data packages. The
data manager organizes all the data from upper-level applications and the data to be relayed. The security manager provides
encryption and decryption to the data manager, routing manager,
and spectrum and channel manager. The knowledge/policy/data
base stores prior knowledge, policies, data, and experiences.
After the nodes are built, a network testbed is ready to be
established.
3) Proposed Network Testbed: Multiple nodes constitute a
network testbed. Fig. 7 shows the proposed network testbed. All
the nodes are connected using gigabit ethernet to a console computer through an ethernet switch. The console computer controls and coordinates all the nodes in the network testbed. This
network testbed can be used not only for cognitive radio, but
also for the smart grid. In smart grid applications, nodes of the
network testbed implement microgrid central controllers, smart
meters, or submeters. Adaptive wireless communications are incorporated into the nodes, and information can be exchanged
between microgrid central controllers, smart meters, and submeters.
4) Security of the Network Testbed: Where there is information, there is a critical need for security. Security is one of the
important issues for the proposed adaptive and reconfigurable
wireless network testbed. The meaning of security is twofold.
First, the data sent out by the nodes should be encrypted, to
prevent unauthorized users from intercepting the data over the
air. However,cryptographic algorithms impose tremendous processing power demands that can be a bottleneck in high-speed
networks [29]. The use of FPGAs for cryptographic applications
has become highly attractive [30], [31]. Cryptographic algorithms will be implemented on the two FPGAs on the proposed
motherboard for the nodes of the network testbed. However, for
the network testbed for smart grid, the optimal choice for the
cryptographic scheme is the topic of ongoing research. Second,
the reconfigurable FPGAs in the network testbed should have

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Support for such work is widely available through both private industry and online communities. Nvidia hosts forums relating to CUDA. GP-you.org is a free online community which
even offers a toolbox called GPUmat, which is geared specifically toward utilizing GPU processing with Matlab. CULA offers more functions than GPUmat, and depending on the license
we acquire, even has private support.
IV. MICROGRID TESTBEDS FOR SMART GRID

IE
W EE
eb P
r
Ve oo
rs f
ion

A microgrid is a localized grouping of electrical sources and


loads. Normally, the microgrid can operate connected to, and
synchronous with the main electrical power grid. However, in
some severe situations, e.g., blackout, electrical outage, low
power quality, and so on, the microgrid can be isolated from
the main electrical power grid [37] and function autonomously.
Compared with the traditional macrogrid, microgrid should
contain distributed generators or distributed energy resources.
Because of independent control strategy and versatile power
sources, microgrid can at least increase the local reliability of
power system, reduce the power loss, maintain the local power
voltage, and enhance power utilization and efficiency.
However, in the current scenario, development of an intelligent microgrid is still in the incipient stages, and there are many
open issues for microgrid from both theory and implementation
considerations. Economic and regulatory issues in microgrid
implementation have been discussed in [38]. An overview of
ongoing research, development, and demonstration projects for
microgrid have been mentioned in [39]. The research activities
in Europe, the United States, and Japan are introduced. Though
the application of distributed energy resources can potentially
reduce the need for traditional power system expansion, controlling a huge number of distributed energy resources leads to
a daunting challenge for operating and controlling the network,
safely and efficiently [39]. The potential role of microgrid in a
wider emerging global electrical energy future has been further
explored in [40]. [40] tried to address the issue whether the microgrid is a viable paradigm for electricity supply expansion.
A. Microgrid Testbeds

Microgrid is a new model for the power grid with a vision towards operating the utility system as efficiently as possible with
connectivity to real-time data through advanced communications. First, integrated renewable or distributed energy sources,
like wind, solar, and microturbines, installed in every single
house can support the local loads without or with less exchange
of power with the main electrical power grid, thereby significantly reducing power loss in the transmission lines. Second,
intelligent communications and efficient power dispatch solutions provide real-time, reliable interaction with consumers and
markets. Third, battery and inverter technology, such as plug-in
vehicles and energy storage, have been evolving such that a
utility can justify the capital cost of the installation based on
the difference in price, buying energy at night at a low price
and selling it back during peak daytime rates. All these technologies enable the microgrid to optimize the use of resources,
secure from threads and hazards, and communicate efficiently
between customers and suppliers. For the purpose of demonstrating and testing the compatibility of intelligent communica-

Fig. 8. Microgrid testbed.

tion methods with renewable energy sources within the microgrid, a microgrid testbed is proposed as shown in Fig. 8. One
lab-based smart grid testbed has been reported in [41]. Intelligent power switch hardware is used to control the behavior of
the power grid testbed. Some functions, such as real-time demand response, price driven demand response, disruption resilience with self-healing as well as flow balance using multiple
path, are demonstrated [41]. However, we will focus more on
the microgrid testbed and try to make the testbed real, versatile,
and comprehensive.
The microgrid under our investigation consists of
smart houses or smart buildings. There is one smart meter installed in each smart house. The smart meter is connected to
the main electrical power grid through the medium voltage or
low voltage power distribution network. In order to implement
distributed control, the interconnection among different smart
meters is assumed. In addition, several submeters in each smart
house are affiliated to the corresponding smart meter. These submeters are used to execute the control signaling to the related
power loads and secondary power sources.
Assume for the smart house , there are
power
loads or power appliances. The power required for power load
is defined as
. means the time when
the energy is consumed by power load
. Here the control
resolution is one hour and the control horizon is one day. Hence,
. Furthermore, five parameters
,
,
,
, and
are specified for power load
.

IEEE TRANSACTIONS ON SMART GRID

is the total energy needed by power load


.
and
are the starting time and ending time, respectively, for power
load
to consume energy.
and
are the minimum
and maximum power levels, respectively, for power load
to
work properly. Thus mathematically speaking,

(1)

IE
W EE
eb P
r
Ve oo
rs f
ion

Assume for the smart house , there are


secondary or renewable power sources such as wind turbine, solar
panel, fuel cell, photovoltaic, and so on. The output power of the
secondary power source is dependent on several factors. The intensity of sunlight has an impact on the output power of solar
panel. The strength and direction of wind can influence the capability of the wind turbine. In addition, there is no dedicated energy storage for each secondary power source. The output power
of the secondary power source is defined as
.
The value of
for any , and can be known by statistical
methods.
Assume for smart house , there is one energy storage. For
load variation and supply intermittency, local energy storage
is a promising option for the microgrid, particularly given the
emergence of new battery technologies [38]. The capacity of
energy storage is
. The charging rate is
, and
the discharging rate is
. Correspondingly, the maxand
imum charging rate and discharging rate are
respectively. Thus,

(2)

Furthermore, for energy storage, the charging and discharging


operation cannot be executed simultaneously,
(3)

The efficiency of charging and discharging is


. The initial energy in energy storage is
which should be less
than or equal to
. Thus,
(4)

and

power from the main electrical power grid when electrical price
is low, and save it in energy storage.
In order to shift energy consumption, and reduce the peak-toaverage ratio (PAR) in the load demand, the concept of real
time electrical price is introduced to the smart grid. The real
time electrical price is a function of time. The ceiling capacity
of power usage can be reduced by charging higher fees during
peak hours [42]. Furthermore, the electrical price is assumed to
be independent of the amount of energy consumption for each
smart house. hence, the real time electrical price is defined as ,
which is given for the following day control. One solution for
predicting real time electrical price has been presented in [43].
The unit of is dollars per kWh.
Besides smart houses, there is one microgrid central controller, one common energy storage, and
common
secondary power sources in the microgrid. These resources are
shared by smart houses within the microgrid. The microgrid
central controller is connected to the main electrical power grid,
the common energy storage, the common secondary power
sources, and smart meters in smart houses. The microgrid central controller together with smart meters takes responsibility
of the operation, maintenance, administration, and provisioning
of microgrid. The point of emphasis for the microgrid central
controller is on the performance of the whole microgrid, while
the smart meters takes more care of the preference for each
smart home.
The interconnection for power flow between different smart
houses is highly complicated, hence, the common energy
storage is very important for the microgrid. The common
energy storage can be treated as energy-exchanging hub within
the microgrid. If the smart house has the surplus energy, this
portion of energy can be sold to the common energy storage at
the selling price
. If smart house has a power shortage,
it can purchase the energy from the common energy storage
instead of the main electrical power grid with the buying price
. All the transactions of energy among smart houses within
the microgrid is controlled by the microgrid central controller.
Auction theory [44] can be explored in the microgrid central
controller to determine the trading price and the trading quantity of energy. Similarly, for the common energy storage, the
capacity
, the charging rate
, the discharging
rate
, the maximum charging rate
, the maximum discharging rate
, and the efficiency
are defined. In addition, the microgrid can also purchase the
power from the main electrical power grid and save it in the
common energy storage.
The output power of the common secondary power sources
will serve the entire community within the microgrid. The
energy generated by the secondary power sources will be fed
into the common energy storage. All smart houses can purchase
this portion of energy from the common energy storage. Similarly, the output power of the common secondary power source
is defined as
.
If power flow between any two smart houses is permitted,
two options are considered based on the basic system model. In
this way, power loss can be reduced. The smart house can sell
the surplus power to its neighbor instead of the common energy
storage which is far away. One option to realize this is to enable

(5)

In one smart house, if the power generated by the secondary


power sources is larger than the power required by the power
loads, the surplus power will be saved in energy storage for the
future usage. If supply is less than demand, the available energy
saved in energy storage will be used to support the operation
of power loads. Additionally, the smart house can purchase the

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

power transaction between the smart houses under the control


of the microgrid central controller. The microgrid central controller uses price as an economic lever to optimize the system
performance. The other option is to implement distributed control in the various smart meters to determine the potential power
transaction. In this manner, power transaction is performed in a
distributed fashion. Though smart meters can cooperate to share
and exchange information, the power transaction decisions by
the smart meter will be self-enforcing.
B. Control Strategy

microgrid with the consideration of reliability, stability, security,


scalability, and so on. Though the real time electrical prices are
dynamic, the price stability should also be taken into account
[48] to avoid the extreme price volatility. In the microgrid, there
are so many restrictions and limitations for the microgrid operation, which can be formulated as the constraints in the optimization issue. Load balancing is the main design requirement.
Electrical generation should be at least matched with power consumption or demand together with power loss in the microgrid.
The carrying capacity of the microgrid, the limitation and efficiency of storage should also be taken into account.
Besides, in order to make the microgrid smart and intelligent, the function of knowledge representation and reasoning
should be added into the control strategy. Knowledge representation and reasoning means the representation of knowledge in
a manner that helps in inferencing from knowledge. Therefore,
control strategy can be augmented by this kind of knowledge
plane that can span vertically over control layers, and horizontally across control modules. There are at least two categories
of functions in knowledge representation and reasoning. One is
a representation of related knowledge. The other is a cognition
loop using artificial intelligence, e.g., machine learning. Prediction is the main function within this framework. Prediction results are very important information for the control strategy to
make a decision beforehand to tackle possible situations in the
future. In this way, the operation of the microgrid will be efficient, smooth, and stable.
In order to solve the specific optimization problem for microgrid control, heuristic algorithms, e.g., particle swarm optimization (PSO), and genetic algorithms, Markov decision process
(MDP), and game theory can be explored. A heuristic algorithm
is an experience-based technique that helps in problem solving.
Heuristic algorithms can give feasible solutions close to the best
answer or the optimal solution. Heuristic algorithm is also suitable for large, complex, nonconvex optimization problems. Coordinated scheduling of residential distributed energy resources
in smart houses has been studied in [49] based on coevolutionary PSO. The coevolutionary PSO with stochastic repulsion
among the particles is proposed to generate the best solutions to
the case study in [49]. Genetic algorithms have also been used to
design optimal placement of hybrid PV-wind systems [50]. Performance improvement and cost minimization are taken into account. Multiobjective genetic algorithms have been exploited to
find the solution to optimum economic and environmental performance problems for small autonomous hybrid power systems
with renewables [51]. Because of conflicting interest between
economic objectives and environmental criteria, the target is to
find a set of nondominated solutions called Pareto-set instead of
a single optimal solution [51].
MDP provides a mathematical framework for modeling decision-making in situations where outcomes are partly random,
and partly under the control of a decision maker. MDP is a discrete time stochastic control process. Thus, MDP can take time
and stochastic factors into the optimization problem, which
can be solved via dynamic programming and reinforcement
learning. Among the renewable resources, wind generation
has most variability and uncertainty, and exhibits multilevel
dynamics across time [52]. Hence, multiple timescale dispatch

IE
W EE
eb P
r
Ve oo
rs f
ion

Though compared with the main electrical power grid, the


microgrid is a small system, a sophisticated control strategy
is still needed to make the microgrid work efficiently. Control
in microgrid refers to the activities, procedures, methods, and
tools that relate to the operation, maintenance, administration,
and provisioning of the microgrid. From the perspective of the
whole microgrid system, central or cooperative control is necessary [45], whereas for the interest of each smart house, distributed or noncooperative control with selfish objective is preferred. Therefore, control strategy for the microgrid should be
layered and hybrid.
From a theoretical point of view, layering as optimization decomposition [46] is one of the general and analytic methodologies for control strategy design. It uses common mathematical
language for thinking, deriving, and comparing [46]. Based on
this mathematical framework, the control strategy relates to the
decomposition scheme of the optimization problem. The overall
optimization problem is decomposed into several subproblems.
Different subproblems will be solved by different functions,
which will be allocated to different control layers and different
control modules, such as the microgrid central controller and
smart meters.
There are two main decompositions, i.e., vertical decomposition and horizontal decomposition [46]. Vertical decomposition
maps an optimization problem into several subproblems which
correspond to different control layers. Different functions
are allocated to different control layers to solve these subproblems. Horizontal decomposition is executed within one
function and decomposes central computation into distributed
computation over geographically different control modules.
Vertical decomposition across the control layers and horizontal
decomposition across the control modules can be conducted
together to decompose the optimization problem systematically
[46]. Furthermore, decomposition structures are not limited
to aforementioned vertical decomposition and horizontal decomposition. Partial decomposition, multilevel decomposition,
and their versatile combinations can lead to many alternative
decompositions [47]. These alternative decompositions can be
exploited as a way to obtain different novel control strategies
[47].
Price-based utility function can be exploited as a design objective for the microgrid. These price-based utility functions at
least cover: 1) prices for purchasing power from the main electric power grid at different times; 2) prices for purchasing power
from the common energy storage or other smart houses within
microgrid; 3) equivalent prices for obtaining power from the
secondary power sources; 4) equivalent prices to maintain the

10

IEEE TRANSACTIONS ON SMART GRID

IE
W EE
eb P
r
Ve oo
rs f
ion

and scheduling for stochastic reliability in the smart grid with


wind generation integration have been studied in [52].
In the microgrid, there will be more than one element,
agent, controller, or decision maker. The control algorithm for
the system with a single agent cannot be well suited for the
distributed control or noncooperative control. How to solve
the multiagent control issue? Game theory gives a general
control methodology to deal with interaction, competition,
and cooperation among decision makers in a complex system.
Game theory is widely used in social sciences, economics,
engineering, and so on. For the smart grid, energy consumption
scheduling issue has been formulated as a game theory problem
[53]. The aim of scheduling is to reduce the total energy cost as
well as PAR in the load demand [53]. Based on the assumption
that the charge for each subscriber is proportional to his/her
total daily load, the energy consumption game can be solved
distributively with minimum exchanging information [53]. In
addition, the unique Nash equilibrium of the energy consumption game is the optimal solution to the central scheduling
problem [53]. The work in [53] has been extended by [54].
Different control strategies based on the degree of information
sharing in the network are studied [54]. Partial knowledge setting and blind setting are considered. The proposed distributed
stochastic energy consumption scheduling algorithms can still
successfully exploit the limited information to improve the
overall load profile [54] In the context of the electrical power
system, auction theory is a popular approach to deal with the
power control issue from an economic point of view. Auction
theory is a kind of game theory, and deals with how agents
act in auction markets. Operation of a multiagent system for
microgrid control has been presented in [55]. Auction theory
has been exploited as the foundation of the proposed algorithm,
the main idea being that every distributed energy resource or
controllable load decides what is best for it, taking into account
the overall benefit [55].
We take one control problem in the single smart house as
an example. Given the real-time electrical price of the main
electrical power grid, the output power profile of the secondary
power source, the power consumption profile of the power load
shown in (1), we would like to see how the capacity of energy storage can affect the cost of purchasing the power from
the main electrical power grid. The key point is that the smart
house can purchase the power from the main electrical power
grid when the electrical price is low, and save the power in
energy storage for future usage. However, there are some constraints for energy storage shown in (3) and (5). Fig. 9 shows the
total cost of purchasing power affected by the capacity of energy
storage. Though the quantitative results shown in Fig. 9 depend
on the simulation setting, the qualitative conclusion is definitive.
The introduction of energy storage can reduce the total cost.
The total cost is decreased as the capacity of energy storage increases. However, when the capacity of energy storage is above
some threshold, the higher capacity energy storage cannot bring
any more benefit. Furthermore, buying, installing, and maintaining the higher capacity energy storage requires more money.
Hence, the energy storage with proper capacity should be exploited to gain benefit for the power consumption in the smart
house.

Fig. 9. The total cost affected by the capacity of energy storage.

C. Security Consideration

Security is as important or more important than any other


performance of interest for the microgrid. To realize a secure
system, security should pervade every aspect of the system design, and be integrated into every system component [56].
For information flow, information security for the microgrid
should include data confidentiality, data authenticity, data
integrity, data freshness, data privacy, public key infrastructure
[57], trusted computing [57], attack detection, attack survivability, intelligent monitoring, cybersecurity, and so on. For
power flow, autonomous recovery is the main security consideration. The microgrid should have the capability of performing
real-time monitoring and quick response. Smart meters together
with many sensors should monitor the operation parameters
or operation status of the microgrid. Furthermore, the unpredictable faults caused by environmental disasters, contrived
attacks, or threats and mechanical failures should be detected
as quickly as possible. False data injection attacks in electricity
markets are under investigation in [58]. The severity of such
attacks is that the attacks can bypass the bad data detection with
the knowledge of the system configuration, and cause profitable financial misconduct [58]. With regards to cyberattack
impact analysis for the smart grid, a graph-theoretic approach
is explored in [59]. This is the unifying and comprehensive
framework to model the interaction and functionality in the
dynamic system, and derive the influence of attack behavior.
Based on the monitored data, the microgrid can adjust itself to
maintain the optimal operation mode, or recover itself to the
correct working condition. Prediction capability is also needed
in the microgrid to ensure its security. The microgrid should
continuously search for hidden troubles or latent dangers in the
system. In addition, the consequences of such latent dangers
will be predicted and evaluated. After analysis, feasible solutions will be given to make the system run in a stable manner.
Isolation or islanding is another way for the microgrid to carry
out autonomous recovery. If the main electrical power grid
breaks down, the microgrid can be isolated and work independently. In this case, the control strategy should be changed and
operation parameters should be reset. Cascade failures cannot
propagate from one system to another [60]. However, the

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

D. Kernel GLRT for Malicious Data Attack


False data injection attack or malicious data attack has been
thoroughly studied in [72]. The attacker with the knowledge
of the power system configuration can launch such an attack
to cause errors in the certain state variables while evading the
existing bad data detection in power system state estimation
[72]. Two papers [73], [74] also discuss malicious data attack
on smart grid state estimation. Both attack strategies and countermeasures are given.
In [73], [74], a detector based on the generalized likelihood
ratio test (GLRT) is introduced. GLRT performs very well when
the large amount of data are available. DC power flow model is

IE
W EE
eb P
r
Ve oo
rs f
ion

seamless transition is of particular importance for the security


consideration [45].
Energy storage cannot only reduce the total cost of purchasing power from the main electrical power grid as shown
in Fig. 9, but also secure the microgrid. The surplus power can
be saved in energy storage temporarily. Otherwise, this portion
of power can have an adverse impact on the quality of power
within the microgrid. Furthermore, secondary power sources
can also gain the benefit of security for the microgrid. Reference [61] studies show how much distributed generators can
mitigate cascading failure in the power grid. The conclusion
is that when only a small number of distributed generators are
used, the likelihood of cascading failure can be dramatically
reduced [61].
Besides, for the security of the microgrid, control or optimization with uncertainty should be taken into account, because
the microgrid is a dynamic complex system, and many of the
variables in the corresponding control or optimization problem
cannot be deterministic or known for sure. There are two basic
approaches to deal with the optimization issue with uncertainty.
One is robust optimization, and the other is stochastic optimization. In robust optimization, the uncertainty model is deterministic and set-based [62]. In contrast, in stochastic optimization,
the uncertainty model is assumed to be random [62]. Robust
optimization, which is a conservative approach [63], can guarantee the performance for all the cases within the set-based uncertainty. In other words, robustness means the performance is
stable with the bounded errors. However, stochastic optimization can only guarantee the performance on average for the uncertainty with known or partially known probability distribution
[63] information. Therefore, there is a trade-off between robustness and performance.
Correspondingly, robust control and stochastic control can be
explored to address the control issue with uncertainty. In this regard, we can take the well studied MDP as an example. When
the transition matrices are uncertain, and the uncertainty is described by possibly nonconvex sets, a robust control problem
for a finite-state, finite-action MDP has been well addressed
in [64]. Furthermore, it is worth noting that partially observable Markov decision process (POMDP) can be treated as one
case in stochastic control for MDP, in which a controller is designed to address the probability of uncertainty in the data. In
POMDP, the observation is not deterministic, and the partial observation brings the uncertainty in terms of observation probability to the controller. MDP can be efficiently solved by a
value iteration algorithm, policy iteration algorithm, or linear
programming. However, solving a POMDP is not easy. The
first detailed algorithm for finding exact solutions for POMDP
have been introduced in [65]. There exists some software tools
for solving POMDP, such as pomdp-solve [66], MADP [67],
ZMDP [68], APPL [69], and Perseus [70]. Besides, for the multiagent system, a single agent POMDP can be extended to multiagent PODMP, interactive POMDP, or decentralized POMDP
[71].

11

(6)

where is the measurement vector; is the system state vector;


is the malicious data vector; is Gaussian measurement noise
vector;
is the deterministic system matrix. We assume at
most measurements are compromised. Under the framework
of GLRT, the distribution of under the two hypotheses differ
only in their means if the state variables are random with a multivariate Gaussian distribution
and multiple measurements are performed under the same [73], [74]

(7)

where
and denotes transpose operator.
Due to the unknown , the GLRT reduces to solve the following optimization problem [73], [74],

(8)

Currently, the kernel method becomes very popular in the society of machine learning. Low dimensional data can be implicitly mapped to the high dimensional feature space by the
kernel trick [75][78] and nonlinear characteristics of data can
be explored implicitly by using different kernels [79]. GLRT
can be extended to kernel GLRT. Kernel GLRT is applied to
kernel matched subspace detector for hyperspectral target detection [79]. Kernel GLRT in the original input space is equivalent to the linear GLRT in the feature space [79].
Assume
, the two hypotheses in the feature
domain corresponding to (7) are

(9)

where is a nonlinear mapping function.


Assume
and
are the estimated covariance matrix
and sample mean of in the feature space, which are given by

(10)

12

IEEE TRANSACTIONS ON SMART GRID

and
(11)
is one sample of and the number of samples is
where
. Thus, the optimization problem (8) can be extended into the
feature space as

IE
W EE
eb P
r
Ve oo
rs f
ion

(12)

the low-rank and sparseness property of the autocovariance matrices of the smart meter signal and wideband interferer, respectively, to effectively separate them prior to ICA processing.
1) Signal Model and Receiver Block Diagram: The system
consists of N smart meters managed by an AP, similar to the
illustration given in [89]. The channel parameters are assumed
to be Rayleigh flat fading in nature, with a large coherence time
indicating a slow time-varying channel. The data transmission
section in the frame is divided into several time slots during
which the active smart meters can simultaneously transmit their
readings. In mathematical form, the data transmission matrix
received by the AP can be expressed as the following ICA signal
model

Based on the kernel trick and the exact derivation in [80], the
objective in the optimization problem (12) can be simplified.
We do not need to know emplicitly. The objective will be the
nonlinear function of
, , and with the
known kernel function. Kernel GLRT opens a new window for
the traditional detection problem. The nonlinear characteristics
of data or the high order correlations among the data can be
exploited by the kernel method, which brings the performance
gain [80].
E. ICA for Recovery of Smart Meter Transmissions in the
Presence of Strong Interference

ICA is a statistical signal processing technique for recovering statistically independent source signals from their linear
mixtures [81][84]. ICA has also been applied to load profile
estimation in electric transmission networks [85]. ICA is very
closely related to the method called blind source separation
(BSS) or blind signal separation [86][88]. The term blind
refers to the fact that we have little or no knowledge about the
system which induces mixing of the source signals.
In this section, the ICA technique is applied for recovering
the simultaneous wireless transmissions of the smart meters installed at each home. The smart meter measures the current load
at each home, and reports that information to the control center
at the power utility station. In order to achieve this, each smart
meter is equipped with a wireless transmitter, and the AP at the
power utility control center collects all the wireless transmissions for processing the information. However, before the transmitted signals can be decoded, it is imperative to separate the
signals received from all the smart meters. In [89], the concept
of compressed sensing [90], [91], was exploited to recover the
sparse smart meter data transmissions by applying the basis pursuit algorithm [92]. However, in [89], it was assumed that the
AP has accurate knowledge of the channel flat fading parameters from the channel estimation period of the data frame. In
this paper, the statistical properties of the signals are employed
to blindly separate the signals using ICA. Therefore, this eliminates the need for estimating the parameters of the channel in
each frame. In this manner, more information can be sent in
each frame. Furthermore, to enhance the security of transmitted
data, the recovery of the smart meter transmissions in the presence of a strong wideband interference or jamming signal is also
considered. In this regard, the recently developed method of robust PCA is used [14], [93]. The robust PCA method exploits

(13)

is the Rayleigh flat fading channel matrix between the meters and the AP, is the pseudorandom spreading code matrix
for the meters, is the source signal matrix transmitted by the
meters, and
is the additive white Gaussian noise (AWGN).
The spreading code is known only to the AP and meters, preventing unauthorized people from accessing and tampering with
the data. Replacing
by the matrix , (13) becomes
(14)

In the context of ICA, is called the mixing matrix. ICA attempts to recover by estimating a matrix that approximates
the inverse of . In this manner, an estimate of the source signal
matrix can be obtained as given by the following equation:

(15)

The main advantage of employing ICA is that it allows the


smart meters to transmit simultaneously, as opposed to the
popular carrier sense multiple access (CSMA) protocol, which
uses a random backoff to avoid collisions in transmissions. This
could result in significant delay in data recovery. In addition,
ICA is a blind process, which means that it does not need
any prior knowledge of the channel or the PN code matrix. As
long as the smart meters transmissions are independent, which
is always the case, since the meters are spatially separated, ICA
can exactly recover all the smart meter signals. However, in the
event of strong interference (14) becomes
(16)

ICA cannot recover the source signals in the presence of


the interferer , since it is not part of the signal mixing model
. As a result,
has to be separated from the observation
matrix , before ICA can be applied. To accomplish this, the
second order statistics of the signal and interferer is exploited.
In particular, the autocovariance function of each row of is
computed. Rewriting (16) in terms of the autocovariance matrices, we obtain [94]
(17)

In (17), is the low-rank autocovariance matrix of the signal


mixture, is the sparse autocovariance matrix of the wideband

13

IE
W EE
eb P
r
Ve oo
rs f
ion

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 10. Robust PCA-ICA based receiver for smart meter data recovery.

interferer consisting of only diagonal entries, and is the autocovariance matrix of the AWGN component. Therefore, (17)
can be written as
(18)

is the power of the interferer, and is the identity


where
matrix. In this manner, (17) exactly fits the robust PCA matrix
model [14]. Therefore, the robust PCA technique can be readily
applied to recover the low-rank signal autocovariance matrix
from the sparse interferer autocovariance matrix. This procedure is repeated for all the rows of the observation matrix .
Therein, once the interferer
is separated from , the signal
model becomes similar to (14), and ICA can be applied to recover the source signals or smart meter transmissions .
The baseband model of the robust PCA-ICA based receiver
(central node or access point) for recovering the simultaneous
transmissions of the smart meters is illustrated in Fig. 10. The
basic receiver functions such as down conversion, analog-todigital conversion, synchronization, etc. are assumed to be completed prior to the data recovery stage in the proposed receiver.
2) Simulation Results: A smart meter network consisting of
meters is assumed, with data transmissions in quadrature phase shift keying (QPSK) modulation format. As a result
of the transmitted data being complex valued, a complex FastICA separation algorithm with a saddle point test called FicaCPLX [13] is used for the blind recovery of source signals.

Since ICA is a block based technique, the processing block


length (number of columns of ) is assume to be 1000 symbols. In the first set of simulations, the performance of the robust PCA-ICA method is studied for different values of
from 1 to 5. The signal-to-noise ratio (SNR) is set at 20 dB.
The signal-to-interference ratio (SIR) is used as the measure of
performance, and is given by the following equation:

SIR

(19)

is the permutation matrix of order N , in our


where
case, a 10 10 matrix.
and
are the absolute
maximum values of the
row, and
columns of , respectively. Ideally, should be a permutation matrix consisting of
only 1s. However, due to the amplitude ambiguity introduced
by the ICA technique, the recovered signals have to scaled accordingly. This can be accomplished by including a small preamble at the beginning of each frame. The SIR (dB) achieved for
different
for QPSK modulation with and without the robust
PCA method is shown in Fig. 11. The corresponding constellation plots for the smart meter 1 signal before and after applying

IEEE TRANSACTIONS ON SMART GRID

IE
W EE
eb P
r
Ve oo
rs f
ion

14

Fig. 11. SIR(dB) versus

for QPSK modulation.

Fig. 12. Scatter plot before applying ICA.

the FicaCPLX algorithm is shown in Figs. 12 and 13, respectively.


In the second set of simulations,
, and the SNR is
varied from 530 dB. For each SNR, the SIR (dB) is plotted in
Fig. 14.
In the third set of simulations,
, and SNR is set at
20 dB. The processing block length is varied between 500 and
1000 symbols. For each block length, the SIR (dB) is plotted in
Fig. 15.
V. CONCLUSION

Roughly speaking, the smart grid has two flows: power flow
and information flow. This paper mainly deals with the second

Fig. 13. Scatter plot after applying ICA.

Fig. 14. SIR (dB) versus SNR (dB) for QPSK modulation.

one. The big picture is to sense, communicate, compute, and


control. This paper takes this picture further by addressing the
system hardware requirements and testbed developments. This
paper is the first to systematically investigate the new idea of
using the next generation wireless technology, cognitive radio
network, for the smart grid. In particular, system architecture,
algorithms, and hardware testbed are studied in detail. A microgrid testbed is also proposed. Control strategies and security
considerations are discussed. The technique of ICA with robust
PCA is applied to smart meter wireless data recovery in the presence of strong wideband interference. When the network testbed
is available, the feasible algorithms for smart grid will be explored in the context of security and reliability.

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

15

[13] Z. Koldovsky and P. Tichavsky, Blind instantaneous noisy mixture


separation with best interference-plus-noise rejection, in Proc. 7th
Int. Conf. Independent Compon. Anal. Signal Separation, 2007, pp.
730737.
[14]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]E. Candes, X. Li, Y.

[15]

[16]

IE
W EE
eb P
r
Ve oo
rs f
ion

[17]

Ma, and J. Wright, Robust principal component analysis, Preprint,


2009.
D. Raychaudhuri, I. Seskar, M. Ott, S. Ganu, K. Ramachandran, H.
Kremo, R. Siracusa, H. Liu, and M. Singh, Overview of the orbit
radio grid testbed for evaluation of next-generation wireless network
protocols, in Proc. IEEE Wireless Commun. Netw. Conf., 2005, pp.
16641669.
I. Broustis, J. Eriksson, S. Krishnamurthy, and M. Faloutsos, A
blueprint for a manageable and affordable wireless testbed: design,
pitfalls and lessons learned, in Proc. 3rd Int. Conf. Testbeds Res.
Infrastruct. Develop. Netw. Communities, 2007, pp. 16.
T. R. Newman, S. S. Hasan, D. Depoy, T. Bose, and J. H. Reed,
Designing and deploying a building-wide cognitive radio network
testbed, IEEE Commun. Mag., vol. 48, no. 9, pp. 106112, 2010.
Ettus Research LLC Sep. 2010 [Online]. Available: http://www.ettus.
com/
GNU Radio Sep. 2010 [Online]. Available: http://www.gnuradio.org/
Z. Chen, N. Guo, Z. Hu, and R. C. Qiu, Experimental validation
of channel state prediction considering delays in practical cognitive
radio, IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 13141325, May
2011.

[18]

[19]
[20]

Fig. 15. SIR (dB) versus block length for QPSK modulation.

[21]

ACKNOWLEDGMENT

The authors gratefully acknowledge Santanu K. Das and


Michael Wicks for their useful discussions.
REFERENCES

[1] M. Shahidehpour, Editorial, IEEE Trans. Smart Grid, vol. 1, no. 1,


p. 1, Jun. 2010.
[2] R. C. Qiu, Cognitive Radio Communications and Networking: Principles and Practice. Hoboken, NJ: Wiley, 2012, p. 600.
[3] R. C. Qiu, M. C. Wicks, L. Li, Z. Hu, and S. J. Hou, Wireless tomography, Part I: A novel approach to remote sensing, in Proc. 5th Int.
Waveform Diversity Design Conf., Niagara Falls, ON, Canada, Aug.
2010.
[4] R. C. Qiu, Z. Hu, M. C. Wicks, L. Li, S. J. Hou, and L. Gary, Wireless
tomography, Part II: A system engineering approach, in Proc. 5th Int.
Waveform Diversity Design Conf., Niagara Falls, ON, Canada, Aug.
2010.
[5] A systems view of the modern grid, N.E.T. Lab., 2007.
[6] S. Amin, Securing the electricity grid, BRIDGE, p. 13, 2010.
[7] V. Sood, D. Fischer, J. Eklund, and T. Brown, Developing a communication infrastructure for the smart grid, in Proc. 2009 IEEE Electr.
Power Energy Conf. (EPEC), pp. 17.
[8]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]R. Qiu, A Cogni-

tive Radio Network Testbed, Proposal to Office of Naval Research


(ONR). : , 2010.
[9] R. C. Qiu, Smart grid research at TTU, Argonne National Laboratory, Feb. 2010 [Online]. Available: http://iweb.tntech.edu/rqiu/publications.htm
[10]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]R. C. Qiu, Cogni-

tive Radio and Smart Grid, IEEE Chapter, Huntsville, Alabama Feb.
18, 2010 [Online]. Available: http://iweb.tntech.edu/rqiu/publications.htm
[11] R. C. Qiu, Z. Chen, N. Guo, Y. Song, P. Zhang, H. Li, and L. Lai, Towards a real-time cognitive radio network testbed: architecture, hardware platform, and application to smart grid, in Proc. 5th IEEE Workshop Netw. Technol. Software-Defined Radio White Space, Jun. 2010.
[12]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]J. Reed, Enabling

Rural Virginias Smart Grid Through White Space Communications. : .

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Lyrtech Incorpo-

rated Small Form Factor SDR Evaluation Module/Development


Platform Users Guide 2010.

[22]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Lyrtech Incorpo-

rated ADACMaster III Users Guide 2009.


[23] Lyrtech Incorporated Sep. 2010 [Online]. Available: http://www.
lyrtech.com/
[24] K. Amiri, Y. Sun, P. Murphy, C. Hunter, J. R. Cavallaro, and A. Sabharwal, WARP, a unified wireless network testbed for education and
research, in Proc. IEEE Int. Conf. Microelectron. Syst. Educ., 2007,
pp. 5354.
[25] Rice University Sep. 2010 [Online]. Available: http://warp.rice.edu/
[26] Mango Communications Sep. 2010 [Online]. Available: http://www.
mangocomm.com/
[27] K. Tan, J. Zhang, J. Fang, H. Liu, Y. Ye, S. Wang, Y. Zhang, H. Wu, W.
Wang, and G. Voelker, Sora: High performance software radio using
general purpose multi-core processors, in Proc. 6th USENIX Symp.
Netw. Syst. Design Implementation USENIX Assoc., 2009, pp. 7590.
[28] K. R. Chowdhury and T. Melodia, Platforms and testbeds for experimental evaluation of cognitive ad hoc networks, IEEE Commun.
Mag., vol. 48, no. 9, pp. 96104, 2010.
[29] V. K. Prasanna and A. Dandalis, FPGA-based cryptography for internet security, in Proc. Online Symp. Electron. Eng., 2000.
[30] T. Wollinger and C. Paar, How secure are FPGAS in cryptographic applications?, in Field-Programmable Logic and Applications. New
York: Springer, 2003, pp. 91100.
[31] F. Crowe, A. Daly, T. Kerins, and W. Marnane, Single-chip FPGA
implementation of a cryptographic co-processor, in Proc. IEEE Int.
Conf. Field-Programm. Technol., 2004, pp. 279285.
[32] I. Hadzic, S. Udani, and J. M. Smith, FPGA Viruses Univ. Pennsylvania, Tech. Rep, 1999.
[33] T. Wollinger, J. Guajardo, and C. Paar, Security on FPGAs: State-ofthe-art implementations and attacks, ACM Trans. Embedded Comput.
Syst. (TECS), vol. 3, no. 3, pp. 534574, 2004.
[34] S. Drimer, Security for Volatile FPGAs Univ. Cambridge, Tech. Rep,
2009.
[35] J. Glenn-Anderson, Gpu-Based Desktop Supercomputing. Santa
Clara, CA: EnParallel Inc, 2008.
[36]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Nvidias Next Gen-

eration CUDA Compute Architecture: Fermi Nvidia, 2009.


[37] M. Pipattanasomporn, H. Feroze, and S. Rahman, Multi-agent systems in a distributed smart grid: Design and implementation, in Proc.
IEEE PES 2009 Power Syst. Conf. Expo. (PSCE09).

16

IEEE TRANSACTIONS ON SMART GRID

[65] E. Sondik, The optimal control of partially observable Markov processes, Ph.D. dissertation, Stanford Univ., Stanford, CA, 1971.
[66] A. Cassandra, Exact and approximate algorithms for partially observable Markov decision processes, Ph.D. dissertation, Brown Univ.,
Providence, RI, 1998.
[67] M. Spaan and F. Oliehoek, The multiagent decision process toolbox:
Software for decision theoretic planning in multiagent systems, Proc.
MSDM, pp. 107121, May 2008.
[68] T. Smith, Probabilistic Planning for Robotic Exploration, Ph.D. dissertation, Carnegie Mellon Univ., Pittsburgh, PA, 2007.
[69] H. Kurniawati, D. Hsu, and W. Lee, Sarsop: Efficient point-based
pomdp planning by approximating optimally reachable belief spaces,
Proc. Robot.: Sci. Syst., 2008.
[70] M. Spaan and N. Vlassis, Perseus: Randomized point-based value iteration for POMDPs, J. Artif. Intell. Res., vol. 24, pp. 195220, 2005.
[71] F. Oliehoek, M. Spaan, and N. Vlassis, Optimal and approximate
Q-value functions for decentralized POMDPs, J. Artif. Intell. Res.,
vol. 32, no. 1, pp. 289353, 2008.
[72] Y. Liu, M. K. Reiter, and P. Ning, False data injection attacks against
state estimation in electric power grids, in Proc. 16th Conf. Comput.
Commun. Security, 2009, pp. 2132.
[73] O. Kosut, L. Jia, R. J. Thomas, and L. Tong, On malicious data attacks
on power system state estimation, in Proc. Universities Power Eng.
Conf. (UPEC), 2010, pp. 16.
[74] O. Kosut, L. Jia, R. Thomas, and L. Tong, Malicious data attacks on
smart grid state estimation: Attack strategies and countermeasures, in
Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg, MD,
Oct. 2010, pp. 220225.
[75] G. Baudat and F. Anouar, Kernel-based methods and function approximation, in Proc. Neural Netw., Washington, DC, Jul. 2001, vol. 2, pp.
12441249.
[76] G. Wu, E. Y. Chang, and N. Panda, Formulating distance functions
via the kernel trick, in Proc. 11th ACM SIGKDD Int. Conf. Knowl.
Discovery Data Mining, Chicago, IL, Aug. 2005, pp. 703709.
[77] J. Marithoz and S. Bengio, A kernel trick for sequences applied to
text-independent speaker verification systems, Pattern Recognition,
vol. 40, no. 8, pp. 23152324, 2007.
[78] J. Wang, J. Lee, and C. Zhang, Kernel Trick Embedded Gaussian Mixture Model, in Proc. Int. Conf. Algorithmic Learn. Theory, New York,
2003, vol. 2842, Lecture Notes in Artificial Intelligence, pp. 159174.
[79] H. Kwon and N. M. Nasrabadi, Kernel matched subspace detectors
for hyperspectral target detection, IEEE Trans. Pattern Anal. Mach.
Intell., vol. 28, no. 2, pp. 178194, 2006.
[80] H. Kwon and N. M. Nasrabadi, Kernel-based anomaly detection in
hyperspectral imagery, Army Research Lab, Adelphi, MD, 2004.
[81] P. Comon, Independent component analysis, a new concept?, Signal
Process., vol. 36, no. 3, pp. 287314, 1994.
[82] A. Hyvarinen and E. Oja, One-unit learning rules for independent
component analysis, in Proc. Adv. Neural Inf. Process. Syst., 1997,
pp. 480486.
[83] A. Hyvarinen and E. Oja, Independent component analysis: Algorithms and applications, Neural Netw., vol. 13, no. 4-5, pp. 411430,
2000.
[84] A. Hyvarinen, J. Karhunen, and E. Oja, Independent Component Analysis. New York: Wiley, 2001.
[85] H. Liao and D. Niebur, Load profile estimation in electric transmission
networks using independent component analysis, IEEE Trans. Power
Syst., vol. 18, no. 2, pp. 707715, 2003.
[86] D. Pham, Blind separation of instantaneous mixture of sources via an
independent component analysis, IEEE Trans. Signal Process., vol.
44, no. 11, pp. 27682779, 2002.
[87] T. Lee, M. Lewicki, and T. Sejnowski, ICA mixture models for unsupervised classification of non-Gaussian classes and automatic context
switching in blind signal separation, IEEE Trans. Pattern Anal. Mach.
Intell., vol. 22, no. 10, pp. 10781089, 2002.
[88] S. Amari, A. Cichocki, and H. Yang, A new learning algorithm for
blind signal separation, in Proc. Adv. Neural Inf. Process. Syst., 1996,
pp. 757763.
[89] H. Li, M. Rukun, L. Lai, and R. C. Qiu, Compressed meter reading
for delay-sensitive and secure load report in smart grid, in Proc. 1st
IEEE Int. Conf. Smart Grid Commun., 2010.
[90] D. Donoho, Compressed sensing, IEEE Trans. Inf. Theory, vol. 52,
no. 4, pp. 12891306, 2006.
[91] E. Candes, J. Romberg, and T. Tao, Robust uncertainty principles:
Exact signal reconstruction from highly incomplete frequency information, IEEE Trans. Inf. Theory, vol. 52, no. 2, pp. 489509, 2006.

IE
W EE
eb P
r
Ve oo
rs f
ion

[38] C. Marnay, H. Asano, S. Papathanassiou, and G. Strbac, Policymaking


for microgrids, IEEE Power Energy Mag., vol. 6, no. 3, pp. 6677,
2008.
[39] N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, Microgrids,
IEEE Power Energy Mag., vol. 5, no. 4, pp. 7894, Jul.Aug. 2007.
[40] G. Venkataramanan and C. Marnay, A larger role for microgrids,
IEEE Power Energy Mag., vol. 6, no. 3, pp. 7882, 2008.
[41] G. Lu, D. De, and W. Song, Smartgridlab: A laboratory-based smart
grid testbed, in Proc. 1st IEEE Int. Conf. Smart Grid Commun.,
Gaithersburg, MD, Oct. 2010, pp. 143148.
[42] P. McDaniel and S. McLaughlin, Security and privacy challenges in
the smart grid, IEEE Security Privacy, vol. 7, no. 3, pp. 7577, 2009.
[43] A. Mohsenian-Rad and A. Leon-Garcia, Optimal residential load control with price prediction in real-time electricity pricing environments,
IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 120133, Sep. 2010.
[44] V. Krishna, Auction Theory. New York: Academic, 2009.
[45] A. Tsikalakis and N. Hatziargyriou, Centralized control for optimizing
microgrids operation, IEEE Trans. Energy Convers., vol. 23, no. 1, pp.
241248, 2008.
[46] M. Chiang, S. H. Low, A. R. Calderbank, and J. C. Doyle, Layering
as optimization decomposition: A mathematical theory of network architectures, Proc. IEEE, vol. 95, no. 1, pp. 255312, 2007.
[47] D. Palomar and M. Chiang, Alternative distributed algorithms for network utility maximization: Framework and applications, IEEE Trans.
Autom. Control, vol. 52, no. 12, pp. 22542269, 2007.
[48] M. Roozbehani, M. Dahleh, and S. Mitter, Dynamic pricing and stabilization of supply and demand in modern electric power grids, in
Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg, MD,
Oct. 2010, pp. 543548.
[49] M. Pedrasa, T. Spooner, and I. MacGill, Coordinated scheduling of
residential distributed energy resources to optimize smart home energy
services, IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 134143, 2010.
[50] M. Masoum, M. Badejani, and M. Kalantar, Optimal placement of hybrid PV-wind systems using genetic algorithm, in Proc. Innov. Smart
Grid Technol. (ISGT) 2010, pp. 15.
[51] Y. A. Katsigiannis, P. S. Georgilakis, and E. S. Karapidakis, Multiobjective genetic algorithm solution to the optimum economic and environmental performance problem of small autonomous hybrid power
systems with renewables, IET Renewable Power Gener., vol. 4, pp.
404419, 2010.
[52] M. He, S. Murugesan, and J. Zhang, Multiple timescale dispatch and
scheduling for stochastic reliability in smart grids with wind generation
integration, Arxiv Preprint , 2010, ArXiv:1008.3932.
[53] A. H. Mohsenian-Rad, V. W. S. Wong, J. Jatskevich, and R. Schober,
Optimal and autonomous incentive-based energy consumption scheduling algorithm for smart grid, in Proc. Innov. Smart Grid Technol.
(ISGT), 2010, pp. 16.
[54] S. Caron and G. Kesidis, Incentive-based energy consumption scheduling algorithms for the smart grid, in Proc. 1st IEEE Int. Conf. Smart
Grid Commun., Gaithersburg, MD, Oct. 2010, pp. 391396.
[55] A. Dimeas and N. Hatziargyriou, Operation of a multiagent system
for microgrid control, IEEE Trans. Power Syst., vol. 20, no. 3, pp.
14471455, 2005.
[56] A. Perrig, J. Stankovic, and D. Wagner, Security in wireless sensor
networks, Commun. ACM, vol. 47, no. 6, pp. 5357, 2004.
[57] A. R. Metke and R. L. Ekl, Security technology for smart grid networks, IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 99107, 2010.
[58] L. Xie, Y. Mo, and B. Sinopoli, False data injection attacks in electricity markets, in Proc. 1st IEEE Int. Conf. Smart Grid Commun.,
Gaithersburg, MD, Oct. 2010, pp. 226231.
[59] D. Kundur, X. Feng, S. Liu, T. Zourntos, and K. Butler-Purry, Towards a framework for cyber attack impact analysis of the electric smart
grid, in Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg,
MD, Oct. 2010, pp. 244249.
[60] H. Khurana, M. Hadley, N. Lu, and D. A. Frincke, Smart-grid security
issues, IEEE Security Privacy, vol. 8, no. 1, pp. 8185, 2010.
[61] X. Chen, H. Dinh, and B. Wang, Cascading failures in smart
gridbenefits of distributed generation, in Proc. 1st IEEE Int. Conf.
Smart Grid Commun., Gaithersburg, MD, Oct. 2010, pp. 7378.
[62] A. Ben-Tal, L. E. Ghaoui, and A. Nemirovski, Robust Optimization.
Princeton, NJ: Princeton Univ. Press, 2009.
[63] P. Setoodeh and S. Haykin, Robust transmit power control for cognitive radio, Proc. IEEE, vol. 97, no. 5, pp. 915939, 2009.
[64] A. Nilim and L. E. Ghaoui, Robust Markov decision processes with
uncertain transition matrices, Oper. Res., vol. 53, no. 5, pp. 780798,
2005.

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

[92] S. Chen, D. Donoho, and M. Saunders, Atomic decomposition by


basis pursuit, SIAM Rev., vol. 43, no. 1, pp. 129159, 2001.
[93]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Z. Lin, M. Chen, L.

Wu, and Y. Ma, The augmented lagrange multiplier method for exact
recovery of corrupted low-rank matrices, Math. Program., 2009.
[94] R. C. Qiu and S. J. Hou, Experimental results on matrix separation algorithm, Cognitive Radio Inst., Tennessee Tech University,
Cookeville, TN, CRI-TR-2011-100, 2011.

he worked on the design and implementation of the physical layer of the first
personal handy-phone system integrated circuit (IC) in China. From 2004 to
2007, he was a Research Engineer with STMicroelectronics, Shanghai, where
he mainly worked on the research of image processing and the development of
prototypes of audio video coding standard (AVS) video decoders. He was one
of the three engineers who developed the first prototype of the AVS1.0 standard
definition real-time video decoder. In 2008, he was a Senior System Engineer
with Huaya Microelectronics Inc, Shanghai, where he worked on the development of the set-top box (STB) and the next-generation STB IC. His research
interests include signal processing and cognitive radio.

Nan Guo (S96-M99-SM10) received the M.S.


degree in telecommunications engineering from the
Beijing University of Posts and Telecommunications, Beijing, China, in 1990 and the Ph.D. degree
in communications and electronic systems from the
University of Electronic Science and Technology of
China (UESTC), Chengdu, China, in 1997.
In September 1990, he became a faculty member
with UESTC. In January 1997, he joined the Center
for Wireless Communications, University of California, San Diego. From December 1999 to January
2002, he was a Research/System Engineer with Golden Bridge Technology,
Inc., West Long Branch, NJ, where he was deeply involved in third-generation
code-division multiple-access system design, intellectual property development, and standardization activities. From June 2002 to February 2003, he was
a Research Engineer with the System Group, Ansoft Corporation, Elmwood
Park, NJ, where his major responsibility was software development, with
emphasis on functionality modeling of emerging technologies. Since 2004, he
has been with the Center for Manufacturing Research, Tennessee Technological
University, Cookeville, doing research and development (R&D) and laboratory
work. He has more than 15 years of industrial and academic experience in
R&D, teaching, and laboratory work. His research interests include wireless
communications, statistic signal processing, optimization and its applications,
and implementation impact on system performance.

IE
W EE
eb P
r
Ve oo
rs f
ion

Robert Caiming Qiu (S93-M96-SM01) received


the Ph.D. degree in electrical engineering from New
York University (former Polytechnic University,
Brooklyn, NY).
He was Founder-CEO and President of Wiscom
Technologies, Inc., manufacturing and marketing
WCDMA chipsets. Wiscom was sold to Intel in
2003. Prior to Wiscom, he worked for GTE Labs,
Inc. (now Verizon), Waltham, MA, and Bell Labs,
Lucent, Whippany, NJ. He is currently Full Professor in the Department of Electrical and Computer
Engineering, Center for Manufacturing Research, Tennessee Technological
University, Cookeville, where he started as an Associate Professor in 2003
before he became a full Professor in 2008. His current interest is in wireless
communication and networking, machine learning, and smart grid technologies.
He has worked in wireless communications and network, machine learning,
smart grid, digital signal processing, EM scattering, composite absorbing
materials, RF microelectronics, UWB, underwater acoustics, and fiber optics.
He holds over 5 patents in WCDMA and authored over 50 journal papers/book
chapters. He is a coauthor of Cognitive Radio Communication and Networking:
Principles and Practice (Wiley, to be published). He is a Guest Book Editor
forUltra-Wideband (UWB) Wireless Communications (Wiley, 2005), and three
special issues on UWB including the IEEE JOURNAL ON SELECTED AREAS
IN COMMUNICATIONS, IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY,
and IEEE TRANSACTIONS ON SMART GRID. He contributed to 3GPP and
IEEE standards bodies. In 1998 he developed the first three courses on 3G for
Bell Labs researchers. He also served as an Adjunct Professor at Polytechnic
University, Brooklyn, NY.
Dr. Qiu serves as Associate Editor, IEEE TRANSACTIONS ON VEHICULAR
TECHNOLOGY and other international journals. He serves as a Member of TPC
for GLOBECOM, ICC, WCNC, MILCOM, ICUWB, etc. In addition, he served
on the advisory board of the New Jersey Center for Wireless Telecommunications (NJCWT). He is included in Marquis Whos Who in America.

17

Zhen Hu received the B.S. degree from Huazhong


University of Science and Technology, Wuhan,
China, in 2004, the M.S. degree from Southeast
University, Nanjing, China, in 2007, and the Ph.D.
degree in electrical engineering from Tennessee
Technological University, Cookeville, in 2010.
He is currently with the Department of Electrical and Computer Engineering, Center for
Manufacturing Research, Tennessee Technological
University. His research interests are system integration and optimization for wireless communication,
radar, sensing, and power systems.

Zhe Chen received the B.S. degree in telecommunications engineering from Northeastern University,
Shenyang, China, in 2000 and the M.S. degree in
signal and information processing from Hangzhou
Dianzi University, Hangzhou, China, in 2003.
He is currently working toward the Ph.D. degree
with the Department of Electrical and Computer
Engineering, Center forManufacturing Research,
Tennessee Technological University, Cookeville.
From 2003 to 2004, he was an Algorithm Engineer with UTStarcom Inc, Shanghai, China, where

Raghuram Ranganathan (S04-M10) received the


B.E. degree in Telecommunication Engineering from
the R.V. College of Engineering, Bangalore, India, in
2001, and the M.S. and Ph.D. degrees in Electrical
Engineering from the University of Central Florida
(UCF), Orlando, FL, in 2004, and 2008, respectively.
From January 2009 to October 2010, he was a
Postdoctoral Research Associate with the Florida
Energy Systems Consortium and UCF. Since
November 2010, he has been working with the
Cognitive Radio Institute at Tennessee Technological University, Cookeville, TN, as a Research and Development Engineer.
His research interests encompass a number of topics in adaptive digital and
statistical signal processing, and its applications in wireless communications,
blind signal separation, array signal processing, cognitive radio, and smart grid.
Dr. Ranganathan has received a number of awards, including the UCF Graduate Provost Fellowship, UCF Electrical Engineering Departmental Fellowship,
and UCF Graduate Research Fellowship.

Shujie Hou received the B.S. degree from the


Department of Mathematics, Daqing Petroleum
Institute, Daqing, China, in 2006 and the M.S. degree from the Department of Applied Mathematics,
Dalian University of Technology, Dalian, China, in
2008. She is currently working toward the Ph.D. degree with the Department of Electrical and Computer
Engineering, Center for Manufacturing Research,
Tennessee Technological University, Cookeville.

18

IEEE TRANSACTIONS ON SMART GRID

IE
W EE
eb P
r
Ve oo
rs f
ion

Gang Zheng is working toward the Ph.D. degree


in the Department of Electrical and Computer
Engineering, Tennessee Technological University,
Cookeville. His major is power systems.

IEEE TRANSACTIONS ON SMART GRID

Cognitive Radio Network for the Smart Grid:


Experimental System Architecture, Control
Algorithms, Security, and Microgrid Testbed
Robert Caiming Qiu, Senior Member, IEEE, Zhen Hu, Zhe Chen, Nan Guo, Senior Member, IEEE,
Raghuram Ranganathan, Member, IEEE, Shujie Hou, and Gang Zheng

Index TermsCognitive radio, grid control, ICA, microgrid, robust PCA, security, smart grid, testbed.

I. INTRODUCTION

promising? Ultimately, artificial intelligence and machine


learning will become the brain of the cognitive radioarguably any machine. How to put cognition into the cognitive
radio network is paramount at this moment. Experimental
knowledge is essential to dispel the fictional notion about
cognitive radio. What is cognitive radio? Obviously, cognitive
radio includes dynamic spectrum access (DSA) as a special
case. Adaptive systems may be a good alternative acronym
for cognitive systemselusive for some peoplein todays
technology.
The ever increasing demand on remote sensing capabilities
directly conflicts with the accelerating awareness of loss of
spectrum allocation. The FCC recommends spectrum policy
that makes 500 megahertz (MHz) of spectrum newly available
for broadband within 10 years, of which 300 MHz should be
made available for mobile use within five years. The need
for dynamic spectrum access using cognitive radio [2] is real
and immediate. The following paper series [3], [4] attempts to
bring together wireless communications with remote sensing,
especially radio frequency tomography.
The ultimate goal is to have integrated sensing, communication and computation systems, for various applications including the smart grid.

IE
Pr EE
int P
r
Ve oo
rs f
ion

AbstractThis paper systematically investigates the novel idea


of applying the next generation wireless technology, cognitive radio
network, for the smart grid. In particular, system architecture, algorithms, and hardware testbed are studied. A microgrid testbed
supporting both power flow and information flow is also proposed.
Control strategies and security considerations are discussed. Furthermore, the concept of independent component analysis (ICA) in
combination with the robust principal component analysis (PCA)
technique is employed to recover data from the simultaneous smart
meter wireless transmissions in the presence of strong wideband
interference. The performance illustrates the gain of bringing the
state of the art mathematics to smart grid.

N THE FIRST issue of these TRANSACTIONS [1], the editorial points out that In a nutshell, smart grid amounts to providing an Internet Protocol (IP) address to every device that is
connected to the electricity grid. The presence of an IP address
enables a two-way communication mechanism between consumers and providers. Roughly speaking, the smart grid will: 1)
reduce blackouts; 2) promote renewable energy usage; 3) give
families more control over their energy diet. The motivation of
this paper is to address items 2 and 3, by studying the interdependencies of critical infrastructure in power and communications. In particular, a microgrid testbed is studied.
The advent of cognitive radio [2] is inevitable. The Moores
law ruthlessly drives computing faster and cheaper. Software
defined radio reflects this trend. However, the combination
of field programmable gate array (FPGA) and digital signal
processor (DSP) hardware is still too expensive. Is the general purpose computing on graphics processor unit (GPGPU)

Manuscript received October 14, 2010; revised April 11, 2011; accepted June
05, 2011. This work was supported by the National Science Foundation under
Grants ECCS-0901420, ECCS-0821658, and ECCS-0622125, and by the Office of Naval Research through Contracts N00014-07-1-0529 and N00014-11-10006. Paper no. TSG-00166-2010
The authors are with the Cognitive Radio Institute, Department of Electrical and Computer Engineering, Center for Manufacturing Research,
Tennessee Technological University, Cookeville, TN 38505, USA (e-mail:
rqiu@tntech.edu; zhu21@students.tntech.edu; zchen42@students.tntech.edu;
nguo@tntech.edu;
raghu80@gmail.com;
shou42@students.tntech.edu;
gzheng42@students.tntech.edu.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2011.2160101

A. Secure Communications in Smart Grid

Among the five identified key technology areas in smart grid,


the implementation of integrated communications is a foundational need, according to [5]. The smart grid in the near future will be required to accommodate increased demands for improved quality and energy efficiency. Solar and wind farms are
joining in for power generation in a distributed fashion. Appliances will become smart and talk to the control centers for optimum operations. Monitoring, managing, and controlling will
be required at all levels. Prediction of electricity prices, weather,
and social/human activities will be taken into account for optimum control. The addition of these new elements will result in
continuously increasing complexity. In order for different subnetworks or elements to be integrated into the smart grid seamlessly, a communication backbone has to be developed prior to
adding various functions. Hence, the earlier the communication
backbone is determined, the lesser the complications that would
be faced later in building the grid.
Although there have been many discussions on how to build
the communication network for the smart grid, it is desirable
to adopt existing communication standards. However, different

1949-3053/$26.00 2011 IEEE

IEEE TRANSACTIONS ON SMART GRID

simulations cannot answer, but also reveal more practical problems for further research. The capability of cognitive radio enables the smart grid, in many aspects, including security. With
minimal modifications to software, a cognitive radio network
testbed can also be used for the smart grid. Recently, there is a
trend of using TV band white space for the smart grid. Cognitive
radio is the most suitable wireless communications technology
using the white space.
The major concerns on hardware platforms for cognitive
radio networks are computing power and response latency. Our
proposed architecture addresses these two concerns. Cognitive
radio introduces intelligence beyond software defined radio
(SDR), like detection and learning algorithms, which means
cognitive radio requires much more computing power than
SDR. This enhanced computing power and reduced response
latency will enable enhanced (distributed) security.
A critical requirement for the smart grid is to enable accurate recovery of the smart meter wireless transmissions at the
central node or access point (AP). One of the impending challenges in this objective is the robustness of the data recovery in
the presence of strong wideband interference, due to easy access of the wireless data to any attacker, and inadequacy of existing physical layer security measures. In this paper, a novel
approach of applying a complex independent component analysis (ICA) technique [13] to smart meter data recovery is proposed, in combination with the recently developed robust principal component analysis (PCA) algorithm [14] for interference
cancellation and security enhancement.
For the feasible implementation of the power system to support the research on the smart grid, a microgrid testbed is proposed together with the potential system model. The microgrid
testbed will include various distributed energy resources, different power loads or appliances, and control modules. Layered
and hybrid control strategy will be explored in the microgrid to
coordinate power generation and power consumption. In order
to secure the microgrid, information security for the information flow and autonomous recovery for the power flow are the
key issues. Besides, control or optimization with uncertainty is
taken into account, because the microgrid is a dynamic complex
system and many of the variables in the corresponding control
or optimization issue cannot be deterministic or known for sure.
Thus, robust control and stochastic control are explored. Robust
control can guarantee the worst case performance, which will
make the microgrid stable and reliable.

IE
Pr EE
int P
r
Ve oo
rs f
ion

from generic communication requirements, end-to-end security


and reliability are even more critical in smart grid communications [6]. According to the Department of Energy (DoE), one
of the critical challenges and opportunities lying ahead is increasing resilience, robustness, and security of the energy infrastructure. Security in the smart grid may be generalized into
two main areasphysical-layer and cybersecurity. In contrast
to cybersecurity that has been addressed intensively, there is
a large room to improve physical-layer security, especially for
wireless communications. For example, traditional spread spectrum techniques enhance security at the physical layer.
On the other hand, latency for time-sensitive data in the smart
grid has to be carefully considered. According to [7], a tolerance
of a maximum latency of 6 cycles, or 100 ms, can be used as a
benchmark for transmitting time-sensitive data. In other words,
to deliver time-sensitive data, dedicated low-latency communication links are required. The overall smart grid communication
network is a combination of different subnetworks with each
aiming at a certain set of requirements. Cognitive radio in the
white spaces (e.g., unused local TV broadcast spectra) can be
adopted to implement the dedicated low-latency communication links.
B. Related Work

The idea of using cognitive radio in the smart grid appears to


be proposed in the literature, for the first time, in [8][11]. In
particular, one of the three objectives of the submitted proposal
[8] in 2009 is apply the proposed network testbed for the smart
grid. The two-page white paper [12] is undated and was not
brought to our attention (through Qius student) until June 2010.
A cognitive radio network testbed is being built at TTU. Reference [11] is the first paper that captures the overall picture of
this projectthat may last at least for another 34 years. The
scope and the philosophy of this project are reported there.
A lot of future applications are network-centric. The design
of this testbed has taken into account the following applications.
Cognitive networking: What is the cognitive radio?
Smart grid: Cognitive radio enhances the network securitythe central challenge in smart grid.
Radiofrequency tomography imaging: The low-cost network has revolutionized remote sensing, e.g., through-wall
imaging.
C. Main Contributions

In this paper, the application of the fifth generation (5G) wireless technology, namely cognitive radio network, to the next
generation energy grid, the smart grid, has been addressed systemically for the first time.
Since the smart grid is a newly developed research topic,
references and literature on using communication testbeds for
smart grid are very less. In the area of wireless communications, cognitive radio is an emerging technique. The essence
of cognitive radio is the ability of communicating over unused
frequency spectrum adaptively and intelligently. A testbed for
cognitive radio networks will not only answer the questions that

II. COMMUNICATION TESTBEDS FOR SMART GRID

A. Hardware Platforms for Cognitive Radio Networks

There have been some wireless network testbeds, such as the


open access research testbed for next-generation wireless networks (ORBIT) [15] and the wireless testbed developed by University of California, Riverside [16]. Some common features
of those wireless network testbeds are summarized as follows.
First, the nodes in the networks are developed based on computer central processing units (CPUs). Second, the nodes use

IE
Pr EE
int P
r
Ve oo
rs f
ion

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 1. USRP2 with WBX RF daughterboard.

Fig. 2. SFF SDR DP with low-band tunable RF module.

802.11 Wi-Fi network interface cards for wireless communications. These network testbeds may work well for evaluating algorithms, protocols, and network performances for Wi-Fi networks. But they are not suitable for cognitive radio networks,
due to their inherent lack of wide-band frequency agility.
Recently, Virginia Tech developed a testbed for cognitive
radio networks with 48 nodes [17], which is an significant
achievement in this area. Each node consists of three parts:
an Intel Xeon processor-based high-performance server, a
Universal Software Radio Peripheral 2 (USRP2), and a custom
developed radiofrequency (RF) daughterboard that covers a
continuous frequency range from 100 MHz to 4 GHz with
variable instantaneous bandwidths from 10 kHz to 20 MHz.
The node is easily capable of frequency agility. However, as
the authors mentioned, the drawbacks of the node are twofold.
First, it is not a low-power processing platform. Second, it is
not capable of mobility.
Regardless of the kind of cognitive radio network testbed, it
is composed of multiple nodes. There exist some commercial
off-the-shelf hardware platforms designed for SDR that may be
used for building the nodes for cognitive radio networks.
1) Universal Software Radio Peripheral 2: USRP and
USRP2 provided by Ettus Research are widely used hardware
platforms in the area of SDR and cognitive radio. USRP2 is
the second generation of USRP, and it became available in
2009 [18]. USRP2 consists of a motherboard, and one or more
selectable RF daughterboards, as shown in Fig. 1.
The major computation power on the motherboard comes
from a Xilinx Spartan-3 XC3S2000 FPGA. The motherboard is
also equipped with a 100 MSPS 14-bit dual channel analog-todigital converter (ADC), a 400 MSPS 16-bit dual channel digital-to-analog converter (DAC), and a gigabit ethernet port that
can be connected to a host computer. There are some RF daughterboards available for USRP2. Among them, a newly developed RF daughterboard called WBX covers a wide frequency
band of 50 MHz to 2.2 GHz, with a nominal noise figure of 57
dB.
Signals are received and down-converted by USRP2, and
its RF daughterboard. Subsequently, they are sent to a host
computer for further processing through the gigabit ethernet.

Most of the processing work is done by the host computer. Data


to be transmitted are sent from the host computer to USRP2
through the same gigabit ethernet, before they are up-converted
and transmitted by USRP2 and its RF daughterboard.
A major advantage of USRP2 is that it works with GNU
radio [19], a open source software with plenty of resources for
SDR and a lot of users, which simplifies and eases the usage
of USRP2. On the other hand, USRP2 is not perfect. First,
the gigabit Ethernet connecting USRP2 and its host computer
introduces random time delays. The operating system on the
host computer may also introduce random time delays. According to our measurement, the response delay of USRP2 is in
the range of several milliseconds to tens of milliseconds [20].
Such random response delay may be acceptable for half-duplex communications. However, in cognitive radio networks,
full-duplex communications are desired and random response
delays may deteriorate the performance of cognitive radio
networks. Second, USRP2 is usually used together with GNU
radio that runs on a host computer. When the instantaneous
bandwidth of USRP2 increases, the CPU on the host computer
gets much busier. Therefore, a multicore CPU is desired,
similar to what Virginia Tech has done to their network testbed.
When the instantaneous bandwidth of USPR2 becomes wider,
and the processing tasks on GNU radio becomes much more
complex, a common CPU may not be competent enough for
real-time processing.
2) Small Form Factor Software Defined Radio Development
Platform: The small form factor (SFF) SDR development platform (DP) provided by Lyrtech in collaboration with Texas Instruments (TI) and Xilinx is a self-contained platform consisting
of three separate boards: digital processing module, data conversion module, and RF module, as shown in Fig. 2 [21][23].
The digital processing module is designed based on
TMS320DM6446 system-on-chip (SoC) from TI and Virtex-4
SX35 FPGA from Xilinx. The TMS320DM6446 SoC has
a C64x+ DSP core running at 594 MHz together with an
advanced RISC machine (ARM9) core running at 297 MHz.
The digital processing module also comes with a 10/100 Mbps
Ethernet port. The data conversion module is equipped with a
125 MSPS 14-bit dual channel ADC and a 500 MSPS 16-bit

IEEE TRANSACTIONS ON SMART GRID

IE
Pr EE
int P
r
Ve oo
rs f
ion

Fig. 3. WARP FPGA board with two radio boards.

dual channel DAC. It also has a Xilinx Virtex-4 LX25 FPGA.


The low-band tunable RF module can be configured to have
either 5 MHz or 20 MHz bandwidth with working frequencies
of 2001050 MHz for the transmitter, and 2001000 MHz for
the receiver. The nominal noise figure of this RF module is 5
dB. Other frequency bands may be covered by several other
RF modules.
There are two favorable features of SFF SDR DP for cognitive radio networks. One is that SFF SDR DP is in small form
factor and can be moved easily. The other is that it is capable
of supporting full-duplex communications. However, there are
also two technical drawbacks of using it to build nodes for cognitive radio networks. One drawback is that its computing capacity is fixed, and it is not easy to upgrade to meet the demands
of cognitive radio networks. The other drawback is the response
time delay. According to our measurement, the response delay
of SFF SDR DP is about tens of milliseconds, and the delay is
constant [20]. Such a nontrivial delay is undesirable for cognitive radio networks, since it may deteriorate the performance.
SFF SDR DP can be viewed as an example of independent
hardware platforms, whereas USRP2 is an example of computer-aided hardware platforms. A comparison between the two
hardware platforms has been reported in [11].
3) Wireless Open-Access Research Platform: The wireless
open-access research platform (WARP) developed by Rice University consists of an FPGA board, and one to four radio boards
[24], as shown in Fig. 3. The second generation of the FPGA
board has a Xilinx Virtex-4 FX100 FPGA and a gigabit ethernet port [25], [26]. The FPGA can be used to implement the
physical layer of wireless communications. There are PowerPC
processors embedded in the FX100 FPGA that can be used to
implement media access control (MAC) and network layer. The
radio board incorporates a dual-channel 65 MSPS 14-bit ADC,
and a dual-channel 125 MSPS 16-bit DAC, covering two frequency ranges of 24002500 MHz and 49005875 MHz, with
a bandwidth of up to 40 MHz.
WARP platform is also a small form factor independent hardware platform, which is attractive for building the nodes of cognitive radio networks. The second advantage of using WARP is
that both the physical layer and MAC layer can be implemented

Fig. 4. Sora radio control board.

on one FPGA, which may simplify the board design, compared


to an FPGA + DSP/ARM architecture. Hence, time delays
introduced by the interface between FPGA and DSP/ARM can
be reduced. However, according to [26], the Virtex-4 FPGA on
WARP is not powerful enough to accommodate both transmitter
and receiver functions at the same time. Thus, full-duplex communications desired by cognitive radio networks cannot be implemented using just one WAPR platform.
4) Microsoft Research Software Radio: Microsoft research
has developed a Software radio (Sora) platform [27]. Sora is
composed of a radio control board (RCB), and a selectable RF
board, and it works with a multicore host computer. The RCB
is shown in Fig. 4.
The RCB contains a Xilinx Virtex-5 FPGA, and it interfaces
with a host computer through a peripheral component interconnect express (PCIe) interface at a rate of up to 16.7 Gbps. Actually, RCB is an interface board for transferring digital signals
between the RF board and computer memory. The RF board
can be a WARP radio board. Processing work including physical layer and MAC layer is done on the host computer.
Sora is a computer-aided platform. The main advantage of
using Sora is that it provides a high-throughput interface between RF boards and a host computer. However, since processing work burdens the host computer, the host computer has
to be very powerful to support all the functions running in real
time. On the other hand, multicore programming and debugging
with speedup tricks is not easy. Moreover, implementing fullduplex communications on one host computer is challenging.
Obviously, a host computer (or server) installed with Sora lacks
mobility.
B. Proposed Testbed for Cognitive Radio Networks and Smart
Grid

All of the above four hardware platforms are designed for


SDR. Two of them connect to a host computer where major processing work is done. The other two are stand-alone hardware
platforms. From the aspect of mobility, stand-alone platforms
are preferable for building the nodes of cognitive radio networks, whereas from the aspect of software development, computer-aided hardware platforms are more practical, since soft-

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 5. Architecture of the proposed motherboard for hardware platforms.

IE
Pr EE
int P
r
Ve oo
rs f
ion

ware development and debugging on a host computer is generally easier. In [28], a compromise between the above two kinds
of hardware platforms is suggested. The authors recommend
performing time-critical tasks in the FPGA, and split MAC design with host and FPGA implementations.
However, to the best of our understanding, compared to the
hardware platforms for SDR, the major concerns on hardware
platforms for cognitive radio networks are computing power
and response time delay. Cognitive radio introduces intelligence beyond SDR, like detection and learning algorithms,
which means cognitive radio requires much more computing
power than SDR. A hardware platform with ample and upgradable computing power is desired for building cognitive radio
testbeds. On the other hand, the desired hardware platform
should have minimum response time delay. If the response time
delay is large, the throughput of cognitive radio networks will
seriously degrade. Moreover, full-duplex communications for
the desired hardware platforms is preferable.
Unfortunately, none of the existing off-the-shelf hardware
platforms can meet the above requirements at the same time.
They are originally designed for SDR, instead of cognitive radio
networks. It is imperative to design a new hardware platform for
building the nodes of cognitive radio networks.
1) Proposed Motherboard for Hardware Platforms: We propose an architecture for the motherboard of the proposed hardware platform. Regarding the RF front-end, existing RF boards
from WARP or USRP2 can be reused to interface with the proposed motherboard. Therefore, one proposed motherboard, and
one or more RF boards constitute one proposed hardware platform.
Fig. 5 shows the proposed architecture of the first generation motherboard and its major components. Two powerful
FPGAs, i.e., a Virtex-6 FPGA and a Virtex-5 FX FPGA, are
employed as core components on the motherboard. All the functions for physical layer and MAC layer are implemented on the
two FPGAs, and no external host computer is required. The proposed hardware platform is stand-alone, thus it has good mobility. The Virtex-5 FX FPGA has PowerPC cores that are dedicated for implementing the MAC layer. Physical layer functions
including spectrum sensing are implemented on the two FPGAs.
The Virtex-5 FPGA is used for the transmitting data path, and it
is connected to one or two RF boards as well as a gigabit ethernet
port. The Virtex-6 FPGA is dedicated for the receiving data
path, with connections to one or two RF boards and an extension
port. The extension port can be used to connect with external
boards to gain access to additional computing resources. The
two FPGAs are connected together by a high-throughput low-latency onboard bus. Both of the FPGAs have access to their own
external memories. The use of two FPGAs is a trade-off between
performance and cost.
The proposed motherboard can provide enough and upgradable computing resource for cognitive radio networks. In addition, the time delays between the two FPGAs are trivial. Moreover, full-duplex communications are supported by one proposed motherboard with two or more RF boards.
2) Proposed Functional Architecture for Building Nodes for
Network Testbeds: Based on the proposed motherboards and
off-the-shelf RF boards, nodes for network testbeds can be im-

Fig. 6. Proposed functional architecture for the nodes.

plemented using the following proposed functional architecture,


as shown in Fig. 6.
The hardware abstraction layer (HAL) is a packaged interface
for upper-level functions that screens hardware-specific details.
It provides data interfaces to both receiving data path and transmitting data path, as well as an access interface to other hardware-specific resources on the hardware platform. The spectrum
and channel manager manages all the spectrum and channel related resources, including links, frequencies, and modulation
methods. There are several functional modules interfaced with
the spectrum and channel manager. The spectrum detection and
prediction module provides the information regarding the availability of some frequency bands. The decision making module
utilizes decision algorithms to make decisions such as which
channel will be used, and when it will be used. More learning al-

IEEE TRANSACTIONS ON SMART GRID

the capability of protecting themselves from being invaded or


tampered. In [32], it is pointed out that FPGAs unfortunately
open another door for malicious users to implement the hardware analogue of a computer virus. Moreover, the protection
of the information processed within the FPGAs is discussed in
[33], [34]. For the smart grid, how do the nodes in the microgrids
prevent their information from being invaded or tampered? This
is an open question for further research.
III. HOST COMPUTER WITH GRAPHICS PROCESSING UNIT AS
COMPUTING ENGINE

IE
Pr EE
int P
r
Ve oo
rs f
ion

The testbed capability can be greatly expanded if a computing


engine is attached. The computing engine can be a host computer which undertakes many tasks including high-computation
oriented tasks. Recently, a computing enhancement technology
called GPGPU emerged in the PC industry. GPGPU refers to a
relatively new method by which the various cores of a graphics
processor unit (GPU) can be utilized for general purpose parallel computing [35]. This idea of utilizing GPUs for nongraphical applications first became popular in 2003, but was limited
by the amount of knowledge required to successfully write such
programs. November 2008 saw the introduction of Nvidias G80
architecture which brought greater versatility through support
of the C computer language, and a more generalized and programmer friendly hardware structure [36].
CUDA is both a hardware and software architecture by
Nvidia, which is actually what allows GPUs to run programs
that have been written using C, C++, Fortran, etc. It works by
executing a kernel across several parallel threads [36]. GPUmat
allows standard MATLAB code to run on GPUs. Furthermore,
CULA is a linear algebra library which has been designed
to utilize the NVidia CUDA architecture for computational
acceleration. This library is designed in a manner such that
those with little or no GPGPU programming experience can
take advantage of the parallel computing power offered by
GPGPU. Given the flexibility offered by this product, it would
be advantageous to consider it for use with the test-bed.
The CULA library is compatible with Python, C/C++, Fortran, and Matlab. When using C/C++, the library is designed
in such a way that the user may simply replace existing functions in the program with those from the library. CULA is designed such that it automatically handles the memory allocation
required with GPGPU programming. This is the most attractive feature of this software because it allows users who are not
experienced with GPGPU programming to take advantage of
the increased speed it offers. The code is also flexible enough
so that more experienced programmers can manually adjust the
memory allocation for the GPU. Such features are very valuable
to our lab because they allow a diverse set of people to work with
the GPUs, while still retaining the option for more advanced
programming. CULA can also use the Matlab MEX compiler
to compile code from within the Matlab environment, which is
capable of utilizing the GPGPU parallel structure. From the descriptions given, this seems to be very similar to the open source
library GPUmat. Our lab has investigated GPUmat in the past
as a means for GPGPU programming, but it was felt that it still
lacked many of the functions we would require for our work.

Fig. 7. Proposed network testbed.

gorithms can be implemented as an independent module to learn


and reason from the inputs. The geolocation module outputs the
latitude and longitude of the node. The spectrum and channel
manager can use such geolocation information to load prior information about current location from the knowledge/policy/
data base. The routing manager employs routing algorithms to
select the best route for sending and relaying data packages. The
data manager organizes all the data from upper-level applications and the data to be relayed. The security manager provides
encryption and decryption to the data manager, routing manager,
and spectrum and channel manager. The knowledge/policy/data
base stores prior knowledge, policies, data, and experiences.
After the nodes are built, a network testbed is ready to be
established.
3) Proposed Network Testbed: Multiple nodes constitute a
network testbed. Fig. 7 shows the proposed network testbed. All
the nodes are connected using gigabit ethernet to a console computer through an ethernet switch. The console computer controls and coordinates all the nodes in the network testbed. This
network testbed can be used not only for cognitive radio, but
also for the smart grid. In smart grid applications, nodes of the
network testbed implement microgrid central controllers, smart
meters, or submeters. Adaptive wireless communications are incorporated into the nodes, and information can be exchanged
between microgrid central controllers, smart meters, and submeters.
4) Security of the Network Testbed: Where there is information, there is a critical need for security. Security is one of the
important issues for the proposed adaptive and reconfigurable
wireless network testbed. The meaning of security is twofold.
First, the data sent out by the nodes should be encrypted, to
prevent unauthorized users from intercepting the data over the
air. However,cryptographic algorithms impose tremendous processing power demands that can be a bottleneck in high-speed
networks [29]. The use of FPGAs for cryptographic applications
has become highly attractive [30], [31]. Cryptographic algorithms will be implemented on the two FPGAs on the proposed
motherboard for the nodes of the network testbed. However, for
the network testbed for smart grid, the optimal choice for the
cryptographic scheme is the topic of ongoing research. Second,
the reconfigurable FPGAs in the network testbed should have

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Support for such work is widely available through both private industry and online communities. Nvidia hosts forums relating to CUDA. GP-you.org is a free online community which
even offers a toolbox called GPUmat, which is geared specifically toward utilizing GPU processing with Matlab. CULA offers more functions than GPUmat, and depending on the license
we acquire, even has private support.
IV. MICROGRID TESTBEDS FOR SMART GRID

IE
Pr EE
int P
r
Ve oo
rs f
ion

A microgrid is a localized grouping of electrical sources and


loads. Normally, the microgrid can operate connected to, and
synchronous with the main electrical power grid. However, in
some severe situations, e.g., blackout, electrical outage, low
power quality, and so on, the microgrid can be isolated from
the main electrical power grid [37] and function autonomously.
Compared with the traditional macrogrid, microgrid should
contain distributed generators or distributed energy resources.
Because of independent control strategy and versatile power
sources, microgrid can at least increase the local reliability of
power system, reduce the power loss, maintain the local power
voltage, and enhance power utilization and efficiency.
However, in the current scenario, development of an intelligent microgrid is still in the incipient stages, and there are many
open issues for microgrid from both theory and implementation
considerations. Economic and regulatory issues in microgrid
implementation have been discussed in [38]. An overview of
ongoing research, development, and demonstration projects for
microgrid have been mentioned in [39]. The research activities
in Europe, the United States, and Japan are introduced. Though
the application of distributed energy resources can potentially
reduce the need for traditional power system expansion, controlling a huge number of distributed energy resources leads to
a daunting challenge for operating and controlling the network,
safely and efficiently [39]. The potential role of microgrid in a
wider emerging global electrical energy future has been further
explored in [40]. [40] tried to address the issue whether the microgrid is a viable paradigm for electricity supply expansion.
A. Microgrid Testbeds

Microgrid is a new model for the power grid with a vision towards operating the utility system as efficiently as possible with
connectivity to real-time data through advanced communications. First, integrated renewable or distributed energy sources,
like wind, solar, and microturbines, installed in every single
house can support the local loads without or with less exchange
of power with the main electrical power grid, thereby significantly reducing power loss in the transmission lines. Second,
intelligent communications and efficient power dispatch solutions provide real-time, reliable interaction with consumers and
markets. Third, battery and inverter technology, such as plug-in
vehicles and energy storage, have been evolving such that a
utility can justify the capital cost of the installation based on
the difference in price, buying energy at night at a low price
and selling it back during peak daytime rates. All these technologies enable the microgrid to optimize the use of resources,
secure from threads and hazards, and communicate efficiently
between customers and suppliers. For the purpose of demonstrating and testing the compatibility of intelligent communica-

Fig. 8. Microgrid testbed.

tion methods with renewable energy sources within the microgrid, a microgrid testbed is proposed as shown in Fig. 8. One
lab-based smart grid testbed has been reported in [41]. Intelligent power switch hardware is used to control the behavior of
the power grid testbed. Some functions, such as real-time demand response, price driven demand response, disruption resilience with self-healing as well as flow balance using multiple
path, are demonstrated [41]. However, we will focus more on
the microgrid testbed and try to make the testbed real, versatile,
and comprehensive.
The microgrid under our investigation consists of
smart houses or smart buildings. There is one smart meter installed in each smart house. The smart meter is connected to
the main electrical power grid through the medium voltage or
low voltage power distribution network. In order to implement
distributed control, the interconnection among different smart
meters is assumed. In addition, several submeters in each smart
house are affiliated to the corresponding smart meter. These submeters are used to execute the control signaling to the related
power loads and secondary power sources.
Assume for the smart house , there are
power
loads or power appliances. The power required for power load
is defined as
. means the time when
the energy is consumed by power load
. Here the control
resolution is one hour and the control horizon is one day. Hence,
. Furthermore, five parameters
,
,
,
, and
are specified for power load
.

IEEE TRANSACTIONS ON SMART GRID

is the total energy needed by power load


.
and
are the starting time and ending time, respectively, for power
load
to consume energy.
and
are the minimum
and maximum power levels, respectively, for power load
to
work properly. Thus mathematically speaking,

(1)

IE
Pr EE
int P
r
Ve oo
rs f
ion

Assume for the smart house , there are


secondary or renewable power sources such as wind turbine, solar
panel, fuel cell, photovoltaic, and so on. The output power of the
secondary power source is dependent on several factors. The intensity of sunlight has an impact on the output power of solar
panel. The strength and direction of wind can influence the capability of the wind turbine. In addition, there is no dedicated energy storage for each secondary power source. The output power
of the secondary power source is defined as
.
The value of
for any , and can be known by statistical
methods.
Assume for smart house , there is one energy storage. For
load variation and supply intermittency, local energy storage
is a promising option for the microgrid, particularly given the
emergence of new battery technologies [38]. The capacity of
energy storage is
. The charging rate is
, and
the discharging rate is
. Correspondingly, the maximum charging rate and discharging rate are
and
respectively. Thus,

(2)

Furthermore, for energy storage, the charging and discharging


operation cannot be executed simultaneously,

(3)

The efficiency of charging and discharging is


. The initial energy in energy storage is
which should be less
than or equal to
. Thus,

(4)

and

power from the main electrical power grid when electrical price
is low, and save it in energy storage.
In order to shift energy consumption, and reduce the peak-toaverage ratio (PAR) in the load demand, the concept of real
time electrical price is introduced to the smart grid. The real
time electrical price is a function of time. The ceiling capacity
of power usage can be reduced by charging higher fees during
peak hours [42]. Furthermore, the electrical price is assumed to
be independent of the amount of energy consumption for each
smart house. hence, the real time electrical price is defined as ,
which is given for the following day control. One solution for
predicting real time electrical price has been presented in [43].
The unit of is dollars per kWh.
Besides smart houses, there is one microgrid central controller, one common energy storage, and
common
secondary power sources in the microgrid. These resources are
shared by smart houses within the microgrid. The microgrid
central controller is connected to the main electrical power grid,
the common energy storage, the common secondary power
sources, and smart meters in smart houses. The microgrid central controller together with smart meters takes responsibility
of the operation, maintenance, administration, and provisioning
of microgrid. The point of emphasis for the microgrid central
controller is on the performance of the whole microgrid, while
the smart meters takes more care of the preference for each
smart home.
The interconnection for power flow between different smart
houses is highly complicated, hence, the common energy
storage is very important for the microgrid. The common
energy storage can be treated as energy-exchanging hub within
the microgrid. If the smart house has the surplus energy, this
portion of energy can be sold to the common energy storage at
the selling price
. If smart house has a power shortage,
it can purchase the energy from the common energy storage
instead of the main electrical power grid with the buying price
. All the transactions of energy among smart houses within
the microgrid is controlled by the microgrid central controller.
Auction theory [44] can be explored in the microgrid central
controller to determine the trading price and the trading quantity of energy. Similarly, for the common energy storage, the
capacity
, the charging rate
, the discharging
rate
, the maximum charging rate
, the maximum discharging rate
, and the efficiency
are defined. In addition, the microgrid can also purchase the
power from the main electrical power grid and save it in the
common energy storage.
The output power of the common secondary power sources
will serve the entire community within the microgrid. The
energy generated by the secondary power sources will be fed
into the common energy storage. All smart houses can purchase
this portion of energy from the common energy storage. Similarly, the output power of the common secondary power source
is defined as
.
If power flow between any two smart houses is permitted,
two options are considered based on the basic system model. In
this way, power loss can be reduced. The smart house can sell
the surplus power to its neighbor instead of the common energy
storage which is far away. One option to realize this is to enable

(5)

In one smart house, if the power generated by the secondary


power sources is larger than the power required by the power
loads, the surplus power will be saved in energy storage for the
future usage. If supply is less than demand, the available energy
saved in energy storage will be used to support the operation
of power loads. Additionally, the smart house can purchase the

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

power transaction between the smart houses under the control


of the microgrid central controller. The microgrid central controller uses price as an economic lever to optimize the system
performance. The other option is to implement distributed control in the various smart meters to determine the potential power
transaction. In this manner, power transaction is performed in a
distributed fashion. Though smart meters can cooperate to share
and exchange information, the power transaction decisions by
the smart meter will be self-enforcing.
B. Control Strategy

microgrid with the consideration of reliability, stability, security,


scalability, and so on. Though the real time electrical prices are
dynamic, the price stability should also be taken into account
[48] to avoid the extreme price volatility. In the microgrid, there
are so many restrictions and limitations for the microgrid operation, which can be formulated as the constraints in the optimization issue. Load balancing is the main design requirement.
Electrical generation should be at least matched with power consumption or demand together with power loss in the microgrid.
The carrying capacity of the microgrid, the limitation and efficiency of storage should also be taken into account.
Besides, in order to make the microgrid smart and intelligent, the function of knowledge representation and reasoning
should be added into the control strategy. Knowledge representation and reasoning means the representation of knowledge in
a manner that helps in inferencing from knowledge. Therefore,
control strategy can be augmented by this kind of knowledge
plane that can span vertically over control layers, and horizontally across control modules. There are at least two categories
of functions in knowledge representation and reasoning. One is
a representation of related knowledge. The other is a cognition
loop using artificial intelligence, e.g., machine learning. Prediction is the main function within this framework. Prediction results are very important information for the control strategy to
make a decision beforehand to tackle possible situations in the
future. In this way, the operation of the microgrid will be efficient, smooth, and stable.
In order to solve the specific optimization problem for microgrid control, heuristic algorithms, e.g., particle swarm optimization (PSO), and genetic algorithms, Markov decision process
(MDP), and game theory can be explored. A heuristic algorithm
is an experience-based technique that helps in problem solving.
Heuristic algorithms can give feasible solutions close to the best
answer or the optimal solution. Heuristic algorithm is also suitable for large, complex, nonconvex optimization problems. Coordinated scheduling of residential distributed energy resources
in smart houses has been studied in [49] based on coevolutionary PSO. The coevolutionary PSO with stochastic repulsion
among the particles is proposed to generate the best solutions to
the case study in [49]. Genetic algorithms have also been used to
design optimal placement of hybrid PV-wind systems [50]. Performance improvement and cost minimization are taken into account. Multiobjective genetic algorithms have been exploited to
find the solution to optimum economic and environmental performance problems for small autonomous hybrid power systems
with renewables [51]. Because of conflicting interest between
economic objectives and environmental criteria, the target is to
find a set of nondominated solutions called Pareto-set instead of
a single optimal solution [51].
MDP provides a mathematical framework for modeling decision-making in situations where outcomes are partly random,
and partly under the control of a decision maker. MDP is a discrete time stochastic control process. Thus, MDP can take time
and stochastic factors into the optimization problem, which
can be solved via dynamic programming and reinforcement
learning. Among the renewable resources, wind generation
has most variability and uncertainty, and exhibits multilevel
dynamics across time [52]. Hence, multiple timescale dispatch

IE
Pr EE
int P
r
Ve oo
rs f
ion

Though compared with the main electrical power grid, the


microgrid is a small system, a sophisticated control strategy
is still needed to make the microgrid work efficiently. Control
in microgrid refers to the activities, procedures, methods, and
tools that relate to the operation, maintenance, administration,
and provisioning of the microgrid. From the perspective of the
whole microgrid system, central or cooperative control is necessary [45], whereas for the interest of each smart house, distributed or noncooperative control with selfish objective is preferred. Therefore, control strategy for the microgrid should be
layered and hybrid.
From a theoretical point of view, layering as optimization decomposition [46] is one of the general and analytic methodologies for control strategy design. It uses common mathematical
language for thinking, deriving, and comparing [46]. Based on
this mathematical framework, the control strategy relates to the
decomposition scheme of the optimization problem. The overall
optimization problem is decomposed into several subproblems.
Different subproblems will be solved by different functions,
which will be allocated to different control layers and different
control modules, such as the microgrid central controller and
smart meters.
There are two main decompositions, i.e., vertical decomposition and horizontal decomposition [46]. Vertical decomposition
maps an optimization problem into several subproblems which
correspond to different control layers. Different functions
are allocated to different control layers to solve these subproblems. Horizontal decomposition is executed within one
function and decomposes central computation into distributed
computation over geographically different control modules.
Vertical decomposition across the control layers and horizontal
decomposition across the control modules can be conducted
together to decompose the optimization problem systematically
[46]. Furthermore, decomposition structures are not limited
to aforementioned vertical decomposition and horizontal decomposition. Partial decomposition, multilevel decomposition,
and their versatile combinations can lead to many alternative
decompositions [47]. These alternative decompositions can be
exploited as a way to obtain different novel control strategies
[47].
Price-based utility function can be exploited as a design objective for the microgrid. These price-based utility functions at
least cover: 1) prices for purchasing power from the main electric power grid at different times; 2) prices for purchasing power
from the common energy storage or other smart houses within
microgrid; 3) equivalent prices for obtaining power from the
secondary power sources; 4) equivalent prices to maintain the

10

IEEE TRANSACTIONS ON SMART GRID

IE
Pr EE
int P
r
Ve oo
rs f
ion

and scheduling for stochastic reliability in the smart grid with


wind generation integration have been studied in [52].
In the microgrid, there will be more than one element,
agent, controller, or decision maker. The control algorithm for
the system with a single agent cannot be well suited for the
distributed control or noncooperative control. How to solve
the multiagent control issue? Game theory gives a general
control methodology to deal with interaction, competition,
and cooperation among decision makers in a complex system.
Game theory is widely used in social sciences, economics,
engineering, and so on. For the smart grid, energy consumption
scheduling issue has been formulated as a game theory problem
[53]. The aim of scheduling is to reduce the total energy cost as
well as PAR in the load demand [53]. Based on the assumption
that the charge for each subscriber is proportional to his/her
total daily load, the energy consumption game can be solved
distributively with minimum exchanging information [53]. In
addition, the unique Nash equilibrium of the energy consumption game is the optimal solution to the central scheduling
problem [53]. The work in [53] has been extended by [54].
Different control strategies based on the degree of information
sharing in the network are studied [54]. Partial knowledge setting and blind setting are considered. The proposed distributed
stochastic energy consumption scheduling algorithms can still
successfully exploit the limited information to improve the
overall load profile [54] In the context of the electrical power
system, auction theory is a popular approach to deal with the
power control issue from an economic point of view. Auction
theory is a kind of game theory, and deals with how agents
act in auction markets. Operation of a multiagent system for
microgrid control has been presented in [55]. Auction theory
has been exploited as the foundation of the proposed algorithm,
the main idea being that every distributed energy resource or
controllable load decides what is best for it, taking into account
the overall benefit [55].
We take one control problem in the single smart house as
an example. Given the real-time electrical price of the main
electrical power grid, the output power profile of the secondary
power source, the power consumption profile of the power load
shown in (1), we would like to see how the capacity of energy storage can affect the cost of purchasing the power from
the main electrical power grid. The key point is that the smart
house can purchase the power from the main electrical power
grid when the electrical price is low, and save the power in
energy storage for future usage. However, there are some constraints for energy storage shown in (3) and (5). Fig. 9 shows the
total cost of purchasing power affected by the capacity of energy
storage. Though the quantitative results shown in Fig. 9 depend
on the simulation setting, the qualitative conclusion is definitive.
The introduction of energy storage can reduce the total cost.
The total cost is decreased as the capacity of energy storage increases. However, when the capacity of energy storage is above
some threshold, the higher capacity energy storage cannot bring
any more benefit. Furthermore, buying, installing, and maintaining the higher capacity energy storage requires more money.
Hence, the energy storage with proper capacity should be exploited to gain benefit for the power consumption in the smart
house.

Fig. 9. The total cost affected by the capacity of energy storage.

C. Security Consideration

Security is as important or more important than any other


performance of interest for the microgrid. To realize a secure
system, security should pervade every aspect of the system design, and be integrated into every system component [56].
For information flow, information security for the microgrid
should include data confidentiality, data authenticity, data
integrity, data freshness, data privacy, public key infrastructure
[57], trusted computing [57], attack detection, attack survivability, intelligent monitoring, cybersecurity, and so on. For
power flow, autonomous recovery is the main security consideration. The microgrid should have the capability of performing
real-time monitoring and quick response. Smart meters together
with many sensors should monitor the operation parameters
or operation status of the microgrid. Furthermore, the unpredictable faults caused by environmental disasters, contrived
attacks, or threats and mechanical failures should be detected
as quickly as possible. False data injection attacks in electricity
markets are under investigation in [58]. The severity of such
attacks is that the attacks can bypass the bad data detection with
the knowledge of the system configuration, and cause profitable financial misconduct [58]. With regards to cyberattack
impact analysis for the smart grid, a graph-theoretic approach
is explored in [59]. This is the unifying and comprehensive
framework to model the interaction and functionality in the
dynamic system, and derive the influence of attack behavior.
Based on the monitored data, the microgrid can adjust itself to
maintain the optimal operation mode, or recover itself to the
correct working condition. Prediction capability is also needed
in the microgrid to ensure its security. The microgrid should
continuously search for hidden troubles or latent dangers in the
system. In addition, the consequences of such latent dangers
will be predicted and evaluated. After analysis, feasible solutions will be given to make the system run in a stable manner.
Isolation or islanding is another way for the microgrid to carry
out autonomous recovery. If the main electrical power grid
breaks down, the microgrid can be isolated and work independently. In this case, the control strategy should be changed and
operation parameters should be reset. Cascade failures cannot
propagate from one system to another [60]. However, the

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

D. Kernel GLRT for Malicious Data Attack


False data injection attack or malicious data attack has been
thoroughly studied in [72]. The attacker with the knowledge
of the power system configuration can launch such an attack
to cause errors in the certain state variables while evading the
existing bad data detection in power system state estimation
[72]. Two papers [73], [74] also discuss malicious data attack
on smart grid state estimation. Both attack strategies and countermeasures are given.
In [73], [74], a detector based on the generalized likelihood
ratio test (GLRT) is introduced. GLRT performs very well when
the large amount of data are available. DC power flow model is

IE
Pr EE
int P
r
Ve oo
rs f
ion

seamless transition is of particular importance for the security


consideration [45].
Energy storage cannot only reduce the total cost of purchasing power from the main electrical power grid as shown
in Fig. 9, but also secure the microgrid. The surplus power can
be saved in energy storage temporarily. Otherwise, this portion
of power can have an adverse impact on the quality of power
within the microgrid. Furthermore, secondary power sources
can also gain the benefit of security for the microgrid. Reference [61] studies show how much distributed generators can
mitigate cascading failure in the power grid. The conclusion
is that when only a small number of distributed generators are
used, the likelihood of cascading failure can be dramatically
reduced [61].
Besides, for the security of the microgrid, control or optimization with uncertainty should be taken into account, because
the microgrid is a dynamic complex system, and many of the
variables in the corresponding control or optimization problem
cannot be deterministic or known for sure. There are two basic
approaches to deal with the optimization issue with uncertainty.
One is robust optimization, and the other is stochastic optimization. In robust optimization, the uncertainty model is deterministic and set-based [62]. In contrast, in stochastic optimization,
the uncertainty model is assumed to be random [62]. Robust
optimization, which is a conservative approach [63], can guarantee the performance for all the cases within the set-based uncertainty. In other words, robustness means the performance is
stable with the bounded errors. However, stochastic optimization can only guarantee the performance on average for the uncertainty with known or partially known probability distribution
[63] information. Therefore, there is a trade-off between robustness and performance.
Correspondingly, robust control and stochastic control can be
explored to address the control issue with uncertainty. In this regard, we can take the well studied MDP as an example. When
the transition matrices are uncertain, and the uncertainty is described by possibly nonconvex sets, a robust control problem
for a finite-state, finite-action MDP has been well addressed
in [64]. Furthermore, it is worth noting that partially observable Markov decision process (POMDP) can be treated as one
case in stochastic control for MDP, in which a controller is designed to address the probability of uncertainty in the data. In
POMDP, the observation is not deterministic, and the partial observation brings the uncertainty in terms of observation probability to the controller. MDP can be efficiently solved by a
value iteration algorithm, policy iteration algorithm, or linear
programming. However, solving a POMDP is not easy. The
first detailed algorithm for finding exact solutions for POMDP
have been introduced in [65]. There exists some software tools
for solving POMDP, such as pomdp-solve [66], MADP [67],
ZMDP [68], APPL [69], and Perseus [70]. Besides, for the multiagent system, a single agent POMDP can be extended to multiagent PODMP, interactive POMDP, or decentralized POMDP
[71].

11

(6)

where is the measurement vector; is the system state vector;


is the malicious data vector; is Gaussian measurement noise
vector;
is the deterministic system matrix. We assume at
most measurements are compromised. Under the framework
of GLRT, the distribution of under the two hypotheses differ
only in their means if the state variables are random with a multivariate Gaussian distribution
and multiple measurements are performed under the same [73], [74]

(7)

and denotes transpose operator.


where
Due to the unknown , the GLRT reduces to solve the following optimization problem [73], [74],

(8)

Currently, the kernel method becomes very popular in the society of machine learning. Low dimensional data can be implicitly mapped to the high dimensional feature space by the
kernel trick [75][78] and nonlinear characteristics of data can
be explored implicitly by using different kernels [79]. GLRT
can be extended to kernel GLRT. Kernel GLRT is applied to
kernel matched subspace detector for hyperspectral target detection [79]. Kernel GLRT in the original input space is equivalent to the linear GLRT in the feature space [79].
Assume
, the two hypotheses in the feature
domain corresponding to (7) are

(9)

where is a nonlinear mapping function.


Assume
and
are the estimated covariance matrix
and sample mean of in the feature space, which are given by

(10)

12

IEEE TRANSACTIONS ON SMART GRID

and
(11)
where
is one sample of and the number of samples is
. Thus, the optimization problem (8) can be extended into the
feature space as

IE
Pr EE
int P
r
Ve oo
rs f
ion

(12)

the low-rank and sparseness property of the autocovariance matrices of the smart meter signal and wideband interferer, respectively, to effectively separate them prior to ICA processing.
1) Signal Model and Receiver Block Diagram: The system
consists of N smart meters managed by an AP, similar to the
illustration given in [89]. The channel parameters are assumed
to be Rayleigh flat fading in nature, with a large coherence time
indicating a slow time-varying channel. The data transmission
section in the frame is divided into several time slots during
which the active smart meters can simultaneously transmit their
readings. In mathematical form, the data transmission matrix
received by the AP can be expressed as the following ICA signal
model

Based on the kernel trick and the exact derivation in [80], the
objective in the optimization problem (12) can be simplified.
We do not need to know emplicitly. The objective will be the
nonlinear function of
, , and with the
known kernel function. Kernel GLRT opens a new window for
the traditional detection problem. The nonlinear characteristics
of data or the high order correlations among the data can be
exploited by the kernel method, which brings the performance
gain [80].
E. ICA for Recovery of Smart Meter Transmissions in the
Presence of Strong Interference

ICA is a statistical signal processing technique for recovering statistically independent source signals from their linear
mixtures [81][84]. ICA has also been applied to load profile
estimation in electric transmission networks [85]. ICA is very
closely related to the method called blind source separation
(BSS) or blind signal separation [86][88]. The term blind
refers to the fact that we have little or no knowledge about the
system which induces mixing of the source signals.
In this section, the ICA technique is applied for recovering
the simultaneous wireless transmissions of the smart meters installed at each home. The smart meter measures the current load
at each home, and reports that information to the control center
at the power utility station. In order to achieve this, each smart
meter is equipped with a wireless transmitter, and the AP at the
power utility control center collects all the wireless transmissions for processing the information. However, before the transmitted signals can be decoded, it is imperative to separate the
signals received from all the smart meters. In [89], the concept
of compressed sensing [90], [91], was exploited to recover the
sparse smart meter data transmissions by applying the basis pursuit algorithm [92]. However, in [89], it was assumed that the
AP has accurate knowledge of the channel flat fading parameters from the channel estimation period of the data frame. In
this paper, the statistical properties of the signals are employed
to blindly separate the signals using ICA. Therefore, this eliminates the need for estimating the parameters of the channel in
each frame. In this manner, more information can be sent in
each frame. Furthermore, to enhance the security of transmitted
data, the recovery of the smart meter transmissions in the presence of a strong wideband interference or jamming signal is also
considered. In this regard, the recently developed method of robust PCA is used [14], [93]. The robust PCA method exploits

(13)

is the Rayleigh flat fading channel matrix between the meters and the AP, is the pseudorandom spreading code matrix
for the meters, is the source signal matrix transmitted by the
meters, and
is the additive white Gaussian noise (AWGN).
The spreading code is known only to the AP and meters, preventing unauthorized people from accessing and tampering with
the data. Replacing
by the matrix , (13) becomes
(14)

In the context of ICA, is called the mixing matrix. ICA attempts to recover by estimating a matrix that approximates
the inverse of . In this manner, an estimate of the source signal
matrix can be obtained as given by the following equation:
(15)

The main advantage of employing ICA is that it allows the


smart meters to transmit simultaneously, as opposed to the
popular carrier sense multiple access (CSMA) protocol, which
uses a random backoff to avoid collisions in transmissions. This
could result in significant delay in data recovery. In addition,
ICA is a blind process, which means that it does not need
any prior knowledge of the channel or the PN code matrix. As
long as the smart meters transmissions are independent, which
is always the case, since the meters are spatially separated, ICA
can exactly recover all the smart meter signals. However, in the
event of strong interference (14) becomes
(16)

ICA cannot recover the source signals in the presence of


the interferer , since it is not part of the signal mixing model
. As a result,
has to be separated from the observation
matrix , before ICA can be applied. To accomplish this, the
second order statistics of the signal and interferer is exploited.
In particular, the autocovariance function of each row of is
computed. Rewriting (16) in terms of the autocovariance matrices, we obtain [94]
(17)

In (17), is the low-rank autocovariance matrix of the signal


mixture, is the sparse autocovariance matrix of the wideband

13

IE
Pr EE
int P
r
Ve oo
rs f
ion

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

Fig. 10. Robust PCA-ICA based receiver for smart meter data recovery.

interferer consisting of only diagonal entries, and is the autocovariance matrix of the AWGN component. Therefore, (17)
can be written as
(18)

where
is the power of the interferer, and is the identity
matrix. In this manner, (17) exactly fits the robust PCA matrix
model [14]. Therefore, the robust PCA technique can be readily
applied to recover the low-rank signal autocovariance matrix
from the sparse interferer autocovariance matrix. This procedure is repeated for all the rows of the observation matrix .
Therein, once the interferer
is separated from , the signal
model becomes similar to (14), and ICA can be applied to recover the source signals or smart meter transmissions .
The baseband model of the robust PCA-ICA based receiver
(central node or access point) for recovering the simultaneous
transmissions of the smart meters is illustrated in Fig. 10. The
basic receiver functions such as down conversion, analog-todigital conversion, synchronization, etc. are assumed to be completed prior to the data recovery stage in the proposed receiver.
2) Simulation Results: A smart meter network consisting of
meters is assumed, with data transmissions in quadrature phase shift keying (QPSK) modulation format. As a result
of the transmitted data being complex valued, a complex FastICA separation algorithm with a saddle point test called FicaCPLX [13] is used for the blind recovery of source signals.

Since ICA is a block based technique, the processing block


length (number of columns of ) is assume to be 1000 symbols. In the first set of simulations, the performance of the robust PCA-ICA method is studied for different values of
from 1 to 5. The signal-to-noise ratio (SNR) is set at 20 dB.
The signal-to-interference ratio (SIR) is used as the measure of
performance, and is given by the following equation:

SIR

(19)

where
is the permutation matrix of order N , in our
case, a 10 10 matrix.
and
are the absolute
maximum values of the
row, and
columns of , respectively. Ideally, should be a permutation matrix consisting of
only 1s. However, due to the amplitude ambiguity introduced
by the ICA technique, the recovered signals have to scaled accordingly. This can be accomplished by including a small preamble at the beginning of each frame. The SIR (dB) achieved for
different
for QPSK modulation with and without the robust
PCA method is shown in Fig. 11. The corresponding constellation plots for the smart meter 1 signal before and after applying

IEEE TRANSACTIONS ON SMART GRID

IE
Pr EE
int P
r
Ve oo
rs f
ion

14

Fig. 11. SIR(dB) versus

for QPSK modulation.

Fig. 12. Scatter plot before applying ICA.

the FicaCPLX algorithm is shown in Figs. 12 and 13, respectively.


In the second set of simulations,
, and the SNR is
varied from 530 dB. For each SNR, the SIR (dB) is plotted in
Fig. 14.
In the third set of simulations,
, and SNR is set at
20 dB. The processing block length is varied between 500 and
1000 symbols. For each block length, the SIR (dB) is plotted in
Fig. 15.
V. CONCLUSION
Roughly speaking, the smart grid has two flows: power flow
and information flow. This paper mainly deals with the second

Fig. 13. Scatter plot after applying ICA.

Fig. 14. SIR (dB) versus SNR (dB) for QPSK modulation.

one. The big picture is to sense, communicate, compute, and


control. This paper takes this picture further by addressing the
system hardware requirements and testbed developments. This
paper is the first to systematically investigate the new idea of
using the next generation wireless technology, cognitive radio
network, for the smart grid. In particular, system architecture,
algorithms, and hardware testbed are studied in detail. A microgrid testbed is also proposed. Control strategies and security
considerations are discussed. The technique of ICA with robust
PCA is applied to smart meter wireless data recovery in the presence of strong wideband interference. When the network testbed
is available, the feasible algorithms for smart grid will be explored in the context of security and reliability.

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

15

[13] Z. Koldovsky and P. Tichavsky, Blind instantaneous noisy mixture


separation with best interference-plus-noise rejection, in Proc. 7th
Int. Conf. Independent Compon. Anal. Signal Separation, 2007, pp.
730737.
[14]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]E. Candes, X. Li, Y.

[15]

[16]

IE
Pr EE
int P
r
Ve oo
rs f
ion

[17]

Ma, and J. Wright, Robust principal component analysis, Preprint,


2009.
D. Raychaudhuri, I. Seskar, M. Ott, S. Ganu, K. Ramachandran, H.
Kremo, R. Siracusa, H. Liu, and M. Singh, Overview of the orbit
radio grid testbed for evaluation of next-generation wireless network
protocols, in Proc. IEEE Wireless Commun. Netw. Conf., 2005, pp.
16641669.
I. Broustis, J. Eriksson, S. Krishnamurthy, and M. Faloutsos, A
blueprint for a manageable and affordable wireless testbed: design,
pitfalls and lessons learned, in Proc. 3rd Int. Conf. Testbeds Res.
Infrastruct. Develop. Netw. Communities, 2007, pp. 16.
T. R. Newman, S. S. Hasan, D. Depoy, T. Bose, and J. H. Reed,
Designing and deploying a building-wide cognitive radio network
testbed, IEEE Commun. Mag., vol. 48, no. 9, pp. 106112, 2010.
Ettus Research LLC Sep. 2010 [Online]. Available: http://www.ettus.
com/
GNU Radio Sep. 2010 [Online]. Available: http://www.gnuradio.org/
Z. Chen, N. Guo, Z. Hu, and R. C. Qiu, Experimental validation
of channel state prediction considering delays in practical cognitive
radio, IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 13141325, May
2011.

[18]

[19]
[20]

Fig. 15. SIR (dB) versus block length for QPSK modulation.

[21]

ACKNOWLEDGMENT

The authors gratefully acknowledge Santanu K. Das and


Michael Wicks for their useful discussions.
REFERENCES

[1] M. Shahidehpour, Editorial, IEEE Trans. Smart Grid, vol. 1, no. 1,


p. 1, Jun. 2010.
[2] R. C. Qiu, Cognitive Radio Communications and Networking: Principles and Practice. Hoboken, NJ: Wiley, 2012, p. 600.
[3] R. C. Qiu, M. C. Wicks, L. Li, Z. Hu, and S. J. Hou, Wireless tomography, Part I: A novel approach to remote sensing, in Proc. 5th Int.
Waveform Diversity Design Conf., Niagara Falls, ON, Canada, Aug.
2010.
[4] R. C. Qiu, Z. Hu, M. C. Wicks, L. Li, S. J. Hou, and L. Gary, Wireless
tomography, Part II: A system engineering approach, in Proc. 5th Int.
Waveform Diversity Design Conf., Niagara Falls, ON, Canada, Aug.
2010.
[5] A systems view of the modern grid, N.E.T. Lab., 2007.
[6] S. Amin, Securing the electricity grid, BRIDGE, p. 13, 2010.
[7] V. Sood, D. Fischer, J. Eklund, and T. Brown, Developing a communication infrastructure for the smart grid, in Proc. 2009 IEEE Electr.
Power Energy Conf. (EPEC), pp. 17.
[8]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]R. Qiu, A Cogni-

tive Radio Network Testbed, Proposal to Office of Naval Research


(ONR). : , 2010.
[9] R. C. Qiu, Smart grid research at TTU, Argonne National Laboratory, Feb. 2010 [Online]. Available: http://iweb.tntech.edu/rqiu/publications.htm
[10]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]R. C. Qiu, Cogni-

tive Radio and Smart Grid, IEEE Chapter, Huntsville, Alabama Feb.
18, 2010 [Online]. Available: http://iweb.tntech.edu/rqiu/publications.htm
[11] R. C. Qiu, Z. Chen, N. Guo, Y. Song, P. Zhang, H. Li, and L. Lai, Towards a real-time cognitive radio network testbed: architecture, hardware platform, and application to smart grid, in Proc. 5th IEEE Workshop Netw. Technol. Software-Defined Radio White Space, Jun. 2010.
[12]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]J. Reed, Enabling
Rural Virginias Smart Grid Through White Space Communications. : .

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Lyrtech Incorporated Small Form Factor SDR Evaluation Module/Development
Platform Users Guide 2010.

[22]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Lyrtech Incorpo-

rated ADACMaster III Users Guide 2009.


[23] Lyrtech Incorporated Sep. 2010 [Online]. Available: http://www.
lyrtech.com/
[24] K. Amiri, Y. Sun, P. Murphy, C. Hunter, J. R. Cavallaro, and A. Sabharwal, WARP, a unified wireless network testbed for education and
research, in Proc. IEEE Int. Conf. Microelectron. Syst. Educ., 2007,
pp. 5354.
[25] Rice University Sep. 2010 [Online]. Available: http://warp.rice.edu/
[26] Mango Communications Sep. 2010 [Online]. Available: http://www.
mangocomm.com/
[27] K. Tan, J. Zhang, J. Fang, H. Liu, Y. Ye, S. Wang, Y. Zhang, H. Wu, W.
Wang, and G. Voelker, Sora: High performance software radio using
general purpose multi-core processors, in Proc. 6th USENIX Symp.
Netw. Syst. Design Implementation USENIX Assoc., 2009, pp. 7590.
[28] K. R. Chowdhury and T. Melodia, Platforms and testbeds for experimental evaluation of cognitive ad hoc networks, IEEE Commun.
Mag., vol. 48, no. 9, pp. 96104, 2010.
[29] V. K. Prasanna and A. Dandalis, FPGA-based cryptography for internet security, in Proc. Online Symp. Electron. Eng., 2000.
[30] T. Wollinger and C. Paar, How secure are FPGAS in cryptographic applications?, in Field-Programmable Logic and Applications. New
York: Springer, 2003, pp. 91100.
[31] F. Crowe, A. Daly, T. Kerins, and W. Marnane, Single-chip FPGA
implementation of a cryptographic co-processor, in Proc. IEEE Int.
Conf. Field-Programm. Technol., 2004, pp. 279285.
[32] I. Hadzic, S. Udani, and J. M. Smith, FPGA Viruses Univ. Pennsylvania, Tech. Rep, 1999.
[33] T. Wollinger, J. Guajardo, and C. Paar, Security on FPGAs: State-ofthe-art implementations and attacks, ACM Trans. Embedded Comput.
Syst. (TECS), vol. 3, no. 3, pp. 534574, 2004.
[34] S. Drimer, Security for Volatile FPGAs Univ. Cambridge, Tech. Rep,
2009.
[35] J. Glenn-Anderson, Gpu-Based Desktop Supercomputing. Santa
Clara, CA: EnParallel Inc, 2008.
[36]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Nvidias Next Gen-

eration CUDA Compute Architecture: Fermi Nvidia, 2009.


[37] M. Pipattanasomporn, H. Feroze, and S. Rahman, Multi-agent systems in a distributed smart grid: Design and implementation, in Proc.
IEEE PES 2009 Power Syst. Conf. Expo. (PSCE09).

16

IEEE TRANSACTIONS ON SMART GRID

[65] E. Sondik, The optimal control of partially observable Markov processes, Ph.D. dissertation, Stanford Univ., Stanford, CA, 1971.
[66] A. Cassandra, Exact and approximate algorithms for partially observable Markov decision processes, Ph.D. dissertation, Brown Univ.,
Providence, RI, 1998.
[67] M. Spaan and F. Oliehoek, The multiagent decision process toolbox:
Software for decision theoretic planning in multiagent systems, Proc.
MSDM, pp. 107121, May 2008.
[68] T. Smith, Probabilistic Planning for Robotic Exploration, Ph.D. dissertation, Carnegie Mellon Univ., Pittsburgh, PA, 2007.
[69] H. Kurniawati, D. Hsu, and W. Lee, Sarsop: Efficient point-based
pomdp planning by approximating optimally reachable belief spaces,
Proc. Robot.: Sci. Syst., 2008.
[70] M. Spaan and N. Vlassis, Perseus: Randomized point-based value iteration for POMDPs, J. Artif. Intell. Res., vol. 24, pp. 195220, 2005.
[71] F. Oliehoek, M. Spaan, and N. Vlassis, Optimal and approximate
Q-value functions for decentralized POMDPs, J. Artif. Intell. Res.,
vol. 32, no. 1, pp. 289353, 2008.
[72] Y. Liu, M. K. Reiter, and P. Ning, False data injection attacks against
state estimation in electric power grids, in Proc. 16th Conf. Comput.
Commun. Security, 2009, pp. 2132.
[73] O. Kosut, L. Jia, R. J. Thomas, and L. Tong, On malicious data attacks
on power system state estimation, in Proc. Universities Power Eng.
Conf. (UPEC), 2010, pp. 16.
[74] O. Kosut, L. Jia, R. Thomas, and L. Tong, Malicious data attacks on
smart grid state estimation: Attack strategies and countermeasures, in
Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg, MD,
Oct. 2010, pp. 220225.
[75] G. Baudat and F. Anouar, Kernel-based methods and function approximation, in Proc. Neural Netw., Washington, DC, Jul. 2001, vol. 2, pp.
12441249.
[76] G. Wu, E. Y. Chang, and N. Panda, Formulating distance functions
via the kernel trick, in Proc. 11th ACM SIGKDD Int. Conf. Knowl.
Discovery Data Mining, Chicago, IL, Aug. 2005, pp. 703709.
[77] J. Marithoz and S. Bengio, A kernel trick for sequences applied to
text-independent speaker verification systems, Pattern Recognition,
vol. 40, no. 8, pp. 23152324, 2007.
[78] J. Wang, J. Lee, and C. Zhang, Kernel Trick Embedded Gaussian Mixture Model, in Proc. Int. Conf. Algorithmic Learn. Theory, New York,
2003, vol. 2842, Lecture Notes in Artificial Intelligence, pp. 159174.
[79] H. Kwon and N. M. Nasrabadi, Kernel matched subspace detectors
for hyperspectral target detection, IEEE Trans. Pattern Anal. Mach.
Intell., vol. 28, no. 2, pp. 178194, 2006.
[80] H. Kwon and N. M. Nasrabadi, Kernel-based anomaly detection in
hyperspectral imagery, Army Research Lab, Adelphi, MD, 2004.
[81] P. Comon, Independent component analysis, a new concept?, Signal
Process., vol. 36, no. 3, pp. 287314, 1994.
[82] A. Hyvarinen and E. Oja, One-unit learning rules for independent
component analysis, in Proc. Adv. Neural Inf. Process. Syst., 1997,
pp. 480486.
[83] A. Hyvarinen and E. Oja, Independent component analysis: Algorithms and applications, Neural Netw., vol. 13, no. 4-5, pp. 411430,
2000.
[84] A. Hyvarinen, J. Karhunen, and E. Oja, Independent Component Analysis. New York: Wiley, 2001.
[85] H. Liao and D. Niebur, Load profile estimation in electric transmission
networks using independent component analysis, IEEE Trans. Power
Syst., vol. 18, no. 2, pp. 707715, 2003.
[86] D. Pham, Blind separation of instantaneous mixture of sources via an
independent component analysis, IEEE Trans. Signal Process., vol.
44, no. 11, pp. 27682779, 2002.
[87] T. Lee, M. Lewicki, and T. Sejnowski, ICA mixture models for unsupervised classification of non-Gaussian classes and automatic context
switching in blind signal separation, IEEE Trans. Pattern Anal. Mach.
Intell., vol. 22, no. 10, pp. 10781089, 2002.
[88] S. Amari, A. Cichocki, and H. Yang, A new learning algorithm for
blind signal separation, in Proc. Adv. Neural Inf. Process. Syst., 1996,
pp. 757763.
[89] H. Li, M. Rukun, L. Lai, and R. C. Qiu, Compressed meter reading
for delay-sensitive and secure load report in smart grid, in Proc. 1st
IEEE Int. Conf. Smart Grid Commun., 2010.
[90] D. Donoho, Compressed sensing, IEEE Trans. Inf. Theory, vol. 52,
no. 4, pp. 12891306, 2006.
[91] E. Candes, J. Romberg, and T. Tao, Robust uncertainty principles:
Exact signal reconstruction from highly incomplete frequency information, IEEE Trans. Inf. Theory, vol. 52, no. 2, pp. 489509, 2006.

IE
Pr EE
int P
r
Ve oo
rs f
ion

[38] C. Marnay, H. Asano, S. Papathanassiou, and G. Strbac, Policymaking


for microgrids, IEEE Power Energy Mag., vol. 6, no. 3, pp. 6677,
2008.
[39] N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, Microgrids,
IEEE Power Energy Mag., vol. 5, no. 4, pp. 7894, Jul.Aug. 2007.
[40] G. Venkataramanan and C. Marnay, A larger role for microgrids,
IEEE Power Energy Mag., vol. 6, no. 3, pp. 7882, 2008.
[41] G. Lu, D. De, and W. Song, Smartgridlab: A laboratory-based smart
grid testbed, in Proc. 1st IEEE Int. Conf. Smart Grid Commun.,
Gaithersburg, MD, Oct. 2010, pp. 143148.
[42] P. McDaniel and S. McLaughlin, Security and privacy challenges in
the smart grid, IEEE Security Privacy, vol. 7, no. 3, pp. 7577, 2009.
[43] A. Mohsenian-Rad and A. Leon-Garcia, Optimal residential load control with price prediction in real-time electricity pricing environments,
IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 120133, Sep. 2010.
[44] V. Krishna, Auction Theory. New York: Academic, 2009.
[45] A. Tsikalakis and N. Hatziargyriou, Centralized control for optimizing
microgrids operation, IEEE Trans. Energy Convers., vol. 23, no. 1, pp.
241248, 2008.
[46] M. Chiang, S. H. Low, A. R. Calderbank, and J. C. Doyle, Layering
as optimization decomposition: A mathematical theory of network architectures, Proc. IEEE, vol. 95, no. 1, pp. 255312, 2007.
[47] D. Palomar and M. Chiang, Alternative distributed algorithms for network utility maximization: Framework and applications, IEEE Trans.
Autom. Control, vol. 52, no. 12, pp. 22542269, 2007.
[48] M. Roozbehani, M. Dahleh, and S. Mitter, Dynamic pricing and stabilization of supply and demand in modern electric power grids, in
Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg, MD,
Oct. 2010, pp. 543548.
[49] M. Pedrasa, T. Spooner, and I. MacGill, Coordinated scheduling of
residential distributed energy resources to optimize smart home energy
services, IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 134143, 2010.
[50] M. Masoum, M. Badejani, and M. Kalantar, Optimal placement of hybrid PV-wind systems using genetic algorithm, in Proc. Innov. Smart
Grid Technol. (ISGT) 2010, pp. 15.
[51] Y. A. Katsigiannis, P. S. Georgilakis, and E. S. Karapidakis, Multiobjective genetic algorithm solution to the optimum economic and environmental performance problem of small autonomous hybrid power
systems with renewables, IET Renewable Power Gener., vol. 4, pp.
404419, 2010.
[52] M. He, S. Murugesan, and J. Zhang, Multiple timescale dispatch and
scheduling for stochastic reliability in smart grids with wind generation
integration, Arxiv Preprint , 2010, ArXiv:1008.3932.
[53] A. H. Mohsenian-Rad, V. W. S. Wong, J. Jatskevich, and R. Schober,
Optimal and autonomous incentive-based energy consumption scheduling algorithm for smart grid, in Proc. Innov. Smart Grid Technol.
(ISGT), 2010, pp. 16.
[54] S. Caron and G. Kesidis, Incentive-based energy consumption scheduling algorithms for the smart grid, in Proc. 1st IEEE Int. Conf. Smart
Grid Commun., Gaithersburg, MD, Oct. 2010, pp. 391396.
[55] A. Dimeas and N. Hatziargyriou, Operation of a multiagent system
for microgrid control, IEEE Trans. Power Syst., vol. 20, no. 3, pp.
14471455, 2005.
[56] A. Perrig, J. Stankovic, and D. Wagner, Security in wireless sensor
networks, Commun. ACM, vol. 47, no. 6, pp. 5357, 2004.
[57] A. R. Metke and R. L. Ekl, Security technology for smart grid networks, IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 99107, 2010.
[58] L. Xie, Y. Mo, and B. Sinopoli, False data injection attacks in electricity markets, in Proc. 1st IEEE Int. Conf. Smart Grid Commun.,
Gaithersburg, MD, Oct. 2010, pp. 226231.
[59] D. Kundur, X. Feng, S. Liu, T. Zourntos, and K. Butler-Purry, Towards a framework for cyber attack impact analysis of the electric smart
grid, in Proc. 1st IEEE Int. Conf. Smart Grid Commun., Gaithersburg,
MD, Oct. 2010, pp. 244249.
[60] H. Khurana, M. Hadley, N. Lu, and D. A. Frincke, Smart-grid security
issues, IEEE Security Privacy, vol. 8, no. 1, pp. 8185, 2010.
[61] X. Chen, H. Dinh, and B. Wang, Cascading failures in smart
gridbenefits of distributed generation, in Proc. 1st IEEE Int. Conf.
Smart Grid Commun., Gaithersburg, MD, Oct. 2010, pp. 7378.
[62] A. Ben-Tal, L. E. Ghaoui, and A. Nemirovski, Robust Optimization.
Princeton, NJ: Princeton Univ. Press, 2009.
[63] P. Setoodeh and S. Haykin, Robust transmit power control for cognitive radio, Proc. IEEE, vol. 97, no. 5, pp. 915939, 2009.
[64] A. Nilim and L. E. Ghaoui, Robust Markov decision processes with
uncertain transition matrices, Oper. Res., vol. 53, no. 5, pp. 780798,
2005.

QIU et al.: COGNITIVE RADIO NETWORK FOR THE SMART GRID

[92] S. Chen, D. Donoho, and M. Saunders, Atomic decomposition by


basis pursuit, SIAM Rev., vol. 43, no. 1, pp. 129159, 2001.
[93]

[PLEASE PROVIDE COMPLETE


PUBLICATION DETAILS]Z. Lin, M. Chen, L.

Wu, and Y. Ma, The augmented lagrange multiplier method for exact
recovery of corrupted low-rank matrices, Math. Program., 2009.
[94] R. C. Qiu and S. J. Hou, Experimental results on matrix separation algorithm, Cognitive Radio Inst., Tennessee Tech University,
Cookeville, TN, CRI-TR-2011-100, 2011.

he worked on the design and implementation of the physical layer of the first
personal handy-phone system integrated circuit (IC) in China. From 2004 to
2007, he was a Research Engineer with STMicroelectronics, Shanghai, where
he mainly worked on the research of image processing and the development of
prototypes of audio video coding standard (AVS) video decoders. He was one
of the three engineers who developed the first prototype of the AVS1.0 standard
definition real-time video decoder. In 2008, he was a Senior System Engineer
with Huaya Microelectronics Inc, Shanghai, where he worked on the development of the set-top box (STB) and the next-generation STB IC. His research
interests include signal processing and cognitive radio.

Nan Guo (S96-M99-SM10) received the M.S.


degree in telecommunications engineering from the
Beijing University of Posts and Telecommunications, Beijing, China, in 1990 and the Ph.D. degree
in communications and electronic systems from the
University of Electronic Science and Technology of
China (UESTC), Chengdu, China, in 1997.
In September 1990, he became a faculty member
with UESTC. In January 1997, he joined the Center
for Wireless Communications, University of California, San Diego. From December 1999 to January
2002, he was a Research/System Engineer with Golden Bridge Technology,
Inc., West Long Branch, NJ, where he was deeply involved in third-generation
code-division multiple-access system design, intellectual property development, and standardization activities. From June 2002 to February 2003, he was
a Research Engineer with the System Group, Ansoft Corporation, Elmwood
Park, NJ, where his major responsibility was software development, with
emphasis on functionality modeling of emerging technologies. Since 2004, he
has been with the Center for Manufacturing Research, Tennessee Technological
University, Cookeville, doing research and development (R&D) and laboratory
work. He has more than 15 years of industrial and academic experience in
R&D, teaching, and laboratory work. His research interests include wireless
communications, statistic signal processing, optimization and its applications,
and implementation impact on system performance.

IE
Pr EE
int P
r
Ve oo
rs f
ion

Robert Caiming Qiu (S93-M96-SM01) received


the Ph.D. degree in electrical engineering from New
York University (former Polytechnic University,
Brooklyn, NY).
He was Founder-CEO and President of Wiscom
Technologies, Inc., manufacturing and marketing
WCDMA chipsets. Wiscom was sold to Intel in
2003. Prior to Wiscom, he worked for GTE Labs,
Inc. (now Verizon), Waltham, MA, and Bell Labs,
Lucent, Whippany, NJ. He is currently Full Professor in the Department of Electrical and Computer
Engineering, Center for Manufacturing Research, Tennessee Technological
University, Cookeville, where he started as an Associate Professor in 2003
before he became a full Professor in 2008. His current interest is in wireless
communication and networking, machine learning, and smart grid technologies.
He has worked in wireless communications and network, machine learning,
smart grid, digital signal processing, EM scattering, composite absorbing
materials, RF microelectronics, UWB, underwater acoustics, and fiber optics.
He holds over 5 patents in WCDMA and authored over 50 journal papers/book
chapters. He is a coauthor of Cognitive Radio Communication and Networking:
Principles and Practice (Wiley, to be published). He is a Guest Book Editor
forUltra-Wideband (UWB) Wireless Communications (Wiley, 2005), and three
special issues on UWB including the IEEE JOURNAL ON SELECTED AREAS
IN COMMUNICATIONS, IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY,
and IEEE TRANSACTIONS ON SMART G RID. He contributed to 3GPP and
IEEE standards bodies. In 1998 he developed the first three courses on 3G for
Bell Labs researchers. He also served as an Adjunct Professor at Polytechnic
University, Brooklyn, NY.
Dr. Qiu serves as Associate Editor, IEEE TRANSACTIONS ON VEHICULAR
TECHNOLOGY and other international journals. He serves as a Member of TPC
for GLOBECOM, ICC, WCNC, MILCOM, ICUWB, etc. In addition, he served
on the advisory board of the New Jersey Center for Wireless Telecommunications (NJCWT). He is included in Marquis Whos Who in America.

17

Zhen Hu received the B.S. degree from Huazhong


University of Science and Technology, Wuhan,
China, in 2004, the M.S. degree from Southeast
University, Nanjing, China, in 2007, and the Ph.D.
degree in electrical engineering from Tennessee
Technological University, Cookeville, in 2010.
He is currently with the Department of Electrical and Computer Engineering, Center for
Manufacturing Research, Tennessee Technological
University. His research interests are system integration and optimization for wireless communication,
radar, sensing, and power systems.

Zhe Chen received the B.S. degree in telecommunications engineering from Northeastern University,
Shenyang, China, in 2000 and the M.S. degree in
signal and information processing from Hangzhou
Dianzi University, Hangzhou, China, in 2003.
He is currently working toward the Ph.D. degree
with the Department of Electrical and Computer
Engineering, Center forManufacturing Research,
Tennessee Technological University, Cookeville.
From 2003 to 2004, he was an Algorithm Engineer with UTStarcom Inc, Shanghai, China, where

Raghuram Ranganathan (S04-M10) received the


B.E. degree in Telecommunication Engineering from
the R.V. College of Engineering, Bangalore, India, in
2001, and the M.S. and Ph.D. degrees in Electrical
Engineering from the University of Central Florida
(UCF), Orlando, FL, in 2004, and 2008, respectively.
From January 2009 to October 2010, he was a
Postdoctoral Research Associate with the Florida
Energy Systems Consortium and UCF. Since
November 2010, he has been working with the
Cognitive Radio Institute at Tennessee Technological University, Cookeville, TN, as a Research and Development Engineer.
His research interests encompass a number of topics in adaptive digital and
statistical signal processing, and its applications in wireless communications,
blind signal separation, array signal processing, cognitive radio, and smart grid.
Dr. Ranganathan has received a number of awards, including the UCF Graduate Provost Fellowship, UCF Electrical Engineering Departmental Fellowship,
and UCF Graduate Research Fellowship.

Shujie Hou received the B.S. degree from the


Department of Mathematics, Daqing Petroleum
Institute, Daqing, China, in 2006 and the M.S. degree from the Department of Applied Mathematics,
Dalian University of Technology, Dalian, China, in
2008. She is currently working toward the Ph.D. degree with the Department of Electrical and Computer
Engineering, Center for Manufacturing Research,
Tennessee Technological University, Cookeville.

18

IEEE TRANSACTIONS ON SMART GRID

IE
Pr EE
int P
r
Ve oo
rs f
ion

Gang Zheng is working toward the Ph.D. degree


in the Department of Electrical and Computer
Engineering, Tennessee Technological University,
Cookeville. His major is power systems.